Plastic deformation of minerals and rocks

Reviews in Mineralogy and Geochemistry Volume 51

Plastic Deformation of Minerals and Rocks FOREWORD Volume 51 was prepared for a short course on Plastic Deformation of Minerals and Rocks, sponsored by the Mineralogical Society of America (MSA) in Berkeley, California, December 4 and 5, 2002. Shun-ichiro Karato (Yale University) and HansRudolf Wenk (University of California–Berkeley) organized both the course and this volume, and the National Science Foundation supported the short course with a generous grant through Yale University. Visit MSA’s web site, http://minsocam.org , for any errata that may eventually be posted for this volume and for a listing of other volumes in the Reviews series. Books may be ordered over the internet. Paul H. Ribbe

Virginia Tech Blacksburg, Virginia

PREFACE This volume highlights some of the frontiers in the study of plastic deformation of minerals and rocks. The research into the plastic properties of minerals and rocks had a major peak in late 1960s to early 1970s, largely stimulated by research in the laboratory of D.T. Griggs and his students and associates. It is the same time when the theory of plate tectonics was established and provided a first quantitative theoretical framework for understanding geological processes. The theory of plate tectonics stimulated the study of deformation properties of Earth materials, both in the brittle and the ductile regimes. Many of the foundations of plastic deformation of minerals and rocks were established during this period. Also, new experimental techniques were developed, including deformation apparatus for high-pressure and high-temperature conditions, electron micros-copy study of defects in minerals, and the X-ray technique of deformation fabric analysis. The field benefited greatly from materials science concepts of deformation that were introduced, including the models of point defects and their interaction with dislocations. A summary of progress is given by the volume “Flow and Fracture of Rocks: The Griggs Volume,” published in 1972 by the American Geophysical Union. Since then, the scope of Earth sciences has greatly expanded. Geodynamics became concerned with the Earth's deep interior where seismologists discovered heterogeneities and anisotropy at all scales that were previously thought to be typical of the crust and the upper mantle. Investigations of the solar system documented new mineral phases and rocks far beyond the Earth. Both domains have received a lot of attention from mineralogists (e.g., summarized in MSA’s Reviews in Mineralogy, Volume 36, “Planetary Materials,” and Volume 37, “Ultra-High Pressure Mineralogy”). Most attention was directed towards crystal chemistry and phase relations, yet an understanding of the deformation behavior is essential for interpreting the dynamic geological processes from geological and geophysical observations. This was largely the reason for a rebirth of the study of rock plasticity, leading to new approaches that include experiments at extreme conditions and modeling of deformation behavior based on physical principles. A wide spectrum of communities emerged that need to use information about mineral plasticity, including mineralogy, petrology, structural geology, seismology, geodynamics and engineering. This was the motivation to organize a iii

workshop, in December 2002 in Emeryville, California, to bridge the very diverse disciplines and facilitate communication. The volume that was written for this workshop should help one to become familiar with a notoriously difficult subject, and the various contributions represent some of the important progress that has been achieved. The spectrum is broad. High-resolution tomographic images of Earth’s interior obtained from seismology need to be interpreted on the bases of materials properties to understand their geodynamic significance. Key issues include the influence of deformation on seismic signatures, such as attenuation and anisotropy, and a new generation of experimental and theoretical studies on rock plasticity has contributed to a better understanding. Extensive space exploration has revealed a variety of tectonic styles on planets and their satellites, underlining the uniqueness of the Earth. To understand why plate tectonics is unique to Earth, one needs to understand the physical mechanisms of localization of deformation at various scales and under different physical conditions. Also here important theoretical and experimental studies have been conducted. In both fields, studies on anisotropy and shear localization, large-strain deformation experiments and quantitative modeling are critical, and these have become available only recently. Complicated interplay among chemical reactions (including partial melting) is a key to understand the evolution of Earth. This book contains two chapters on the developments of new techniques of experimental studies: one is large-strain shear deformation (Chapter 1 by Mackwell and Paterson) and another is deformation experiments under ultrahigh pressures (Chapter 2 by Durham et al.). Both technical developments are the results of years of efforts that are opening up new avenues of research along which rich new results are expected to be obtained. Details of physical and chemical processes of deformation in the crust and the upper mantle are much better understood through the combination of well controlled laboratory experiments with observations on “real” rocks deformed in Earth. Chapter 3 by Tullis and Chapter 4 by Hirth address the issues of deformation of crustal rocks and the upper mantle, respectively. In Chapter 5 Kohlstedt reviews the interplay of partial melting and deformation, an important subject in understanding the chemical evolution of Earth. Cordier presents in Chapter 6 an overview of the new results of ultrahigh pressure deformation of deep mantle minerals and discusses microscopic mechanisms controlling the variation of deformation mechanisms with minerals in the deep mantle. Green and Marone review in Chapter 7 the stability of deformation under deep mantle conditions with special reference to phase transformations and their relationship to the origin of intermediate depth and deep-focus earthquakes. In Chapter 8 Schulson provides a detailed description of fracture mechanisms of ice, including the critical brittle-ductile transition that is relevant not only for glaciology, planetology and engineering, but for structural geology as well. In Chapter 9 Cooper provides a review of experimental and theoretical studies on seismic wave attenuation, which is a critical element in interpreting distribution of seismic wave velocities and attenuation. Chapter 10 by Wenk reviews the relationship between crystal preferred orientation and macroscopic anisotropy, illustrating it with case studies. In Chapter 11 Dawson presents recent progress in polycrystal plasticity to model the development of anisotropic fabrics both at the microscopic and macroscopic scale. Such studies form the basis for geodynamic interpretation of seismic anisotropy. Finally, in Chapter 12 Montagner and Guillot present a thorough review of seismic anisotropy of the upper mantle covering the vast regions of geodynamic interests, using a global surface wave data set. In Chapter 13 Bercovici and Karato summarize the theoretical aspects of shear localization. All chapters contain extensive reference lists to guide readers to the more specialized literature. iv

Obviously this book does not cover all the areas related to plastic deformation of minerals and rocks. Important topics that are not fully covered in this book include mechanisms of semi-brittle deformation and the interplay between microstructure evolution and deformation at different levels, such as dislocation substructures and grainsize evolution (“self-organization”). However, we hope that this volume provides a good introduction for graduate students in Earth science or materials science as well as the researchers in these areas to enter this multidisciplinary field. We thank all the authors who spent their precious time in making a great contribution to our community. We also are grateful to the Mineralogical Society of America for sponsoring this book and the accompanying workshop and to the National Science Foundation for generous financial support of the entire effort. Paul Ribbe, the series editor, did a superb job in coordinating the book and managing to unify all the different formats into a coherent volume by delicately pressuring authors, without breaking their enthusiasm. Shun-ichiro Karato Yale University New Haven, Connecticut

Hans-Rudolf Wenk University of California Berkeley, California

December 2002

v

Reviews in Mineralogy and Geochemistry Volume 51

Plastic Deformation of Minerals and Rocks FOREWORD Volume 51 was prepared for a short course on Plastic Deformation of Minerals and Rocks, sponsored by the Mineralogical Society of America (MSA) in Berkeley, California, December 4 and 5, 2002. Shun-ichiro Karato (Yale University) and HansRudolf Wenk (University of California–Berkeley) organized both the course and this volume, and the National Science Foundation supported the short course with a generous grant through Yale University. Visit MSA’s web site, http://minsocam.org , for any errata that may eventually be posted for this volume and for a listing of other volumes in the Reviews series. Books may be ordered over the internet. Paul H. Ribbe

Virginia Tech Blacksburg, Virginia

PREFACE This volume highlights some of the frontiers in the study of plastic deformation of minerals and rocks. The research into the plastic properties of minerals and rocks had a major peak in late 1960s to early 1970s, largely stimulated by research in the laboratory of D.T. Griggs and his students and associates. It is the same time when the theory of plate tectonics was established and provided a first quantitative theoretical framework for understanding geological processes. The theory of plate tectonics stimulated the study of deformation properties of Earth materials, both in the brittle and the ductile regimes. Many of the foundations of plastic deformation of minerals and rocks were established during this period. Also, new experimental techniques were developed, including deformation apparatus for high-pressure and high-temperature conditions, electron micros-copy study of defects in minerals, and the X-ray technique of deformation fabric analysis. The field benefited greatly from materials science concepts of deformation that were introduced, including the models of point defects and their interaction with dislocations. A summary of progress is given by the volume “Flow and Fracture of Rocks: The Griggs Volume,” published in 1972 by the American Geophysical Union. Since then, the scope of Earth sciences has greatly expanded. Geodynamics became concerned with the Earth's deep interior where seismologists discovered heterogeneities and anisotropy at all scales that were previously thought to be typical of the crust and the upper mantle. Investigations of the solar system documented new mineral phases and rocks far beyond the Earth. Both domains have received a lot of attention from mineralogists (e.g., summarized in MSA’s Reviews in Mineralogy, Volume 36, “Planetary Materials,” and Volume 37, “Ultra-High Pressure Mineralogy”). Most attention was directed towards crystal chemistry and phase relations, yet an understanding of the deformation behavior is essential for interpreting the dynamic geological processes from geological and geophysical observations. This was largely the reason for a rebirth of the study of rock plasticity, leading to new approaches that include experiments at extreme conditions and modeling of deformation behavior based on physical principles. A wide spectrum of communities emerged that need to use information about mineral plasticity, including mineralogy, petrology, structural geology, seismology, geodynamics and engineering. This was the motivation to organize a iii

workshop, in December 2002 in Emeryville, California, to bridge the very diverse disciplines and facilitate communication. The volume that was written for this workshop should help one to become familiar with a notoriously difficult subject, and the various contributions represent some of the important progress that has been achieved. The spectrum is broad. High-resolution tomographic images of Earth’s interior obtained from seismology need to be interpreted on the bases of materials properties to understand their geodynamic significance. Key issues include the influence of deformation on seismic signatures, such as attenuation and anisotropy, and a new generation of experimental and theoretical studies on rock plasticity has contributed to a better understanding. Extensive space exploration has revealed a variety of tectonic styles on planets and their satellites, underlining the uniqueness of the Earth. To understand why plate tectonics is unique to Earth, one needs to understand the physical mechanisms of localization of deformation at various scales and under different physical conditions. Also here important theoretical and experimental studies have been conducted. In both fields, studies on anisotropy and shear localization, large-strain deformation experiments and quantitative modeling are critical, and these have become available only recently. Complicated interplay among chemical reactions (including partial melting) is a key to understand the evolution of Earth. This book contains two chapters on the developments of new techniques of experimental studies: one is large-strain shear deformation (Chapter 1 by Mackwell and Paterson) and another is deformation experiments under ultrahigh pressures (Chapter 2 by Durham et al.). Both technical developments are the results of years of efforts that are opening up new avenues of research along which rich new results are expected to be obtained. Details of physical and chemical processes of deformation in the crust and the upper mantle are much better understood through the combination of well controlled laboratory experiments with observations on “real” rocks deformed in Earth. Chapter 3 by Tullis and Chapter 4 by Hirth address the issues of deformation of crustal rocks and the upper mantle, respectively. In Chapter 5 Kohlstedt reviews the interplay of partial melting and deformation, an important subject in understanding the chemical evolution of Earth. Cordier presents in Chapter 6 an overview of the new results of ultrahigh pressure deformation of deep mantle minerals and discusses microscopic mechanisms controlling the variation of deformation mechanisms with minerals in the deep mantle. Green and Marone review in Chapter 7 the stability of deformation under deep mantle conditions with special reference to phase transformations and their relationship to the origin of intermediate depth and deep-focus earthquakes. In Chapter 8 Schulson provides a detailed description of fracture mechanisms of ice, including the critical brittle-ductile transition that is relevant not only for glaciology, planetology and engineering, but for structural geology as well. In Chapter 9 Cooper provides a review of experimental and theoretical studies on seismic wave attenuation, which is a critical element in interpreting distribution of seismic wave velocities and attenuation. Chapter 10 by Wenk reviews the relationship between crystal preferred orientation and macroscopic anisotropy, illustrating it with case studies. In Chapter 11 Dawson presents recent progress in polycrystal plasticity to model the development of anisotropic fabrics both at the microscopic and macroscopic scale. Such studies form the basis for geodynamic interpretation of seismic anisotropy. Finally, in Chapter 12 Montagner and Guillot present a thorough review of seismic anisotropy of the upper mantle covering the vast regions of geodynamic interests, using a global surface wave data set. In Chapter 13 Bercovici and Karato summarize the theoretical aspects of shear localization. All chapters contain extensive reference lists to guide readers to the more specialized literature. iv

Obviously this book does not cover all the areas related to plastic deformation of minerals and rocks. Important topics that are not fully covered in this book include mechanisms of semi-brittle deformation and the interplay between microstructure evolution and deformation at different levels, such as dislocation substructures and grainsize evolution (“self-organization”). However, we hope that this volume provides a good introduction for graduate students in Earth science or materials science as well as the researchers in these areas to enter this multidisciplinary field. We thank all the authors who spent their precious time in making a great contribution to our community. We also are grateful to the Mineralogical Society of America for sponsoring this book and the accompanying workshop and to the National Science Foundation for generous financial support of the entire effort. Paul Ribbe, the series editor, did a superb job in coordinating the book and managing to unify all the different formats into a coherent volume by delicately pressuring authors, without breaking their enthusiasm. Shun-ichiro Karato Yale University New Haven, Connecticut

Hans-Rudolf Wenk University of California Berkeley, California

December 2002

v

Table of Contents  1 New Developments in Deformation Studies: High-Strain Deformation Stephen J. Mackwell, Mervyn S. Paterson INTRODUCTION............................................................................................................................1 EXPERIMENTAL METHODS FOR LARGE STRAINS ..............................................................1 Coaxial deformation.................................................................................................................. 2 Rotational deformation ............................................................................................................. 2 LARGE-STRAIN BEHAVIOR IN ROCKS — GENERAL CONSIDERATIONS........................4 LARGE-STRAIN BEHAVIOR IN ROCKS — CASE STUDIES ..................................................6 Monomineralic crustal rocks..................................................................................................... 6 Monomineralic mantle rocks .................................................................................................... 9 Polyphase systems .................................................................................................................. 12 SUMMARY ...................................................................................................................................13 APPENDIX I: STRESS ANALYSIS FOR DIAGONALLY SPLIT-CYLINDER SHEAR TEST14 APPENDIX II: THE TORSION TEST ..........................................................................................15 REFERENCES...............................................................................................................................17

2 New Developments in Deformation Experiments at High Pressure William B. Durham, Donald J. Weidner, Shun-ichiro Karato, Yanbin Wang INTRODUCTION..........................................................................................................................21 Why is pressure important? .....................................................................................................22 Terminology related to strength and deformation....................................................................23 A BRIEF HISTORY OF HIGH-PRESSURE APPARATUS ........................................................24 To 5 GPa: Cylindrical devices .................................................................................................24 5 GPa and above: Anvil devices ..............................................................................................26 MEASUREMENT METHODS AT HIGH PRESSURE ...............................................................27 Stress measurement..................................................................................................................27 Strain rate measurement...........................................................................................................32 MODERN TECHNIQUES FOR DEFORMATION AT HIGH PRESSURES..............................34 Sample assemblies ...................................................................................................................34 Modifications to the sample assembly in a multianvil press....................................................36 Diamond-anvil cell...................................................................................................................37 Deformation-DIA.....................................................................................................................37 Rotational Drickamer apparatus (RDA) ..................................................................................40 SUMMARY AND PERSPECTIVES.............................................................................................44 ACKNOWLEDGMENTS..............................................................................................................45 REFERENCES...............................................................................................................................46

3

Deformation of Granitic Rocks: Experimental Studies and Natural Examples Jan Tullis

INTRODUCTION..........................................................................................................................51 CRYSTAL PLASTICITY PROCESSES .......................................................................................52 MONOMINERALIC AGGREGATES: QUARTZ........................................................................55 Water........................................................................................................................................55 Slip systems .............................................................................................................................56 Recrystallization mechanisms and microstructures .................................................................56 LPOs ........................................................................................................................................62

Flow laws and piezometers......................................................................................................64 Comparisons with naturally deformed quartzites ....................................................................65 MONOMINERALIC AGGREGATES: FELDSPAR ....................................................................70 Water........................................................................................................................................71 Slip systems .............................................................................................................................71 Recrystallization mechanisms and microstructures .................................................................71 LPOs ........................................................................................................................................74 Flow laws and piezometers......................................................................................................74 Comparisons with naturally deformed feldspars .....................................................................75 POLYPHASE AGGREGATES .....................................................................................................78 Quartz-feldspar aggregates ......................................................................................................78 Effects of mica .........................................................................................................................79 Comparisons with naturally deformed granitic rocks ..............................................................81 APPLICATIONS AND IMPLICATIONS.....................................................................................85 Information from dislocation creep microstructures................................................................85 Flow laws for crustal rocks ......................................................................................................86 Assessing the strength of the crust...........................................................................................88 ACKNOWLEDGMENTS..............................................................................................................89 REFERENCES...............................................................................................................................89

4

Laboratory Constraints on the Rheology of the Upper Mantle Greg Hirth

INTRODUCTION..........................................................................................................................97 BACKGROUND............................................................................................................................97 Brittle deformation and low-temperature plasticity .................................................................97 High-temperature creep ...........................................................................................................99 Brittle-ductile/brittle-plastic transitions .................................................................................102 INSIGHTS, CAVEATS AND QUESTIONS ABOUT APPLYING LABORATORY DATA TO CONSTRAIN THE RHEOLOGY OF THE OCEANIC MANTLE .......................103 Strength of the lithosphere and the depth of oceanic earthquakes .........................................106 Strength of plate boundaries in the viscous regime ...............................................................109 High-temperature creep and the viscosity of the mantle........................................................112 ACKNOWLEDGMENTS............................................................................................................116 REFERENCES.............................................................................................................................116

5

Partial Melting and Deformation David L. Kohlstedt

INTRODUCTION........................................................................................................................121 MELT DISTRIBUTIONS IN NON-DEFORMING ROCKS......................................................121 DEFORMATION OF PARTIALLY MOLTEN ROCKS............................................................127 MELT DISTRIBUTIONS IN DEFORMING ROCKS................................................................131 ACKNOWLEDGMENTS............................................................................................................134 REFERENCES.............................................................................................................................134

6

Dislocations and Slip Systems of Mantle Minerals Patrick Cordier

INTRODUCTION........................................................................................................................137 BASIC CONSIDERATIONS.......................................................................................................137 EXPERIMENTAL ADVANCES.................................................................................................142 High-pressure deformation experiments............................................................................... 143 Transmission electron microscopy ....................................................................................... 144 X-ray diffraction peak broadening........................................................................................ 147 PLASTICITY OF MANTLE PHASES........................................................................................147 SiO2 system........................................................................................................................... 149 Olivine, wadsleyite, ringwoodite .......................................................................................... 153 Garnets .................................................................................................................................. 167 WHERE DO WE STAND?..........................................................................................................171 PERSPECTIVES: FROM ATOMIC TO THE GLOBAL SCALE..............................................173 REFERENCES.............................................................................................................................174

7

Instability of Deformation Harry W. Green, II, Chris Marone

INTRODUCTION........................................................................................................................181 SHEARING INSTABILITY ........................................................................................................182 EXPERIMENTAL HIGH-PRESSURE FAULTING MECHANISMS.......................................185 Dehydration-induced embrittlement ..................................................................................... 185 Transformation-induced faulting .......................................................................................... 188 “Brittle” versus “plastic” shear failure.................................................................................. 192 Thermal runaway due to shear heating ................................................................................. 192 APPLICATION TO EARTHQUAKE MECHANISMS..............................................................193 Earthquake distribution with depth ....................................................................................... 193 Mineral reactions available to trigger earthquakes ............................................................... 194 CONCLUSIONS AND SPECULATIONS ..................................................................................196 REFERENCES.............................................................................................................................197

8

Brittle Failure of Ice Erland M. Schulson

INTRODUCTION........................................................................................................................201 FAILURE UNDER TENSION ....................................................................................................202 Characteristics....................................................................................................................... 202 Failure mechanisms .............................................................................................................. 205 Ductile ice ............................................................................................................................. 206 BRITTLE FAILURE UNDER COMPRESSION ........................................................................207 Overview............................................................................................................................... 207 Brittle compressive failure under multiaxial loading............................................................ 210 MICROSTRUCTURAL FEATURES AND MECHANISMS OF BRITTLE COMPRESSIVE FAILURE ....................................................................................222 Longitudinal splits, material collapse and wing cracks ........................................................ 222 Shear faults and comb-cracks ............................................................................................... 225 DUCTILE-TO-BRITTLE TRANSITION UNDER COMPRESSION ........................................234 Definition .............................................................................................................................. 234 Transition models ................................................................................................................. 235 ON THE FORMATION OF “LEADS” IN THE ARCTIC SEA ICE COVER ...........................239 ICE AND ROCK..........................................................................................................................245 ACKNOWLEDGMENTS............................................................................................................246 REFERENCES.............................................................................................................................246

9

Seismic Wave Attenuation: Energy Dissipation in Viscoelastic Crystalline Solids Reid F. Cooper

INTRODUCTION........................................................................................................................253 LINEAR VISCOELASTICITY: A CHEMICAL KINETICS PERSPECTIVE ..........................253 SPRING AND DASHPOT MODELS OF VISCOELASTICITY ...............................................256 ATTENUATION AND THE LINEAR VISCOELASTIC MODELS.........................................259 THE ATTENUATION BAND/HIGH-TEMPERATURE BACKGROUND ..............................262 ISOLATION/CHARACTERIZATION OF A SINGLE PHYSICAL MECHANISM PRODUCING A POWER-LAW ATTENUATION SPECTRUM: THE INTRINSIC TRANSIENT IN DIFFUSIONAL CREEP..................................................266 DATA EXTRAPOLATION AND APPLICABILITY OF EXPERIMENTS TO GEOPHYSICAL CONDITIONS ...........................................................271 SUBGRAIN ABSORPTION AND THE ATTENUATION BAND............................................276 IMPACT OF PARTIAL MELTING ............................................................................................278 IMPACTS OF DEFECT CHEMISTRY (INCLUDING WATER) AND THE STRUCTURE(S) OF INTERFACES .....................................................................283 FINAL COMMENTS...................................................................................................................286 ACKNOWLEDGMENTS............................................................................................................286 REFERENCES.............................................................................................................................287

10

Texture and Anisotropy Hans-Rudolph Wenk

INTRODUCTION........................................................................................................................291 MEASUREMENTS OF TEXTURES ..........................................................................................291 Overview............................................................................................................................... 291 X-ray pole figure goniometer................................................................................................ 292 Synchrotron X-rays............................................................................................................... 292 Neutron diffraction................................................................................................................ 294 Transmission electron microscope (TEM and HVEM) ........................................................ 295 Scanning electron microscope (SEM)................................................................................... 296 Comparison of methods ........................................................................................................ 296 DATA ANALYSIS ......................................................................................................................297 Orientation distributions ....................................................................................................... 297 From pole figures to ODF..................................................................................................... 298 Use of whole diffraction spectra ........................................................................................... 299 Statistical considerations of single orientation measurements.............................................. 300 From textures to elastic anisotropy ....................................................................................... 301 POLYCRYSTAL PLASTICITY SIMULATIONS......................................................................302 General comments ................................................................................................................ 302 Deformation .......................................................................................................................... 303 Recrystallization ................................................................................................................... 306 APPLICATIONS OF POLYCRYSTAL PLASTICITY ..............................................................307 Introduction........................................................................................................................... 307 Coaxial thinning versus non-coaxial shearing (calcite) ........................................................ 307 Anisotropy in the upper mantle (olivine).............................................................................. 309 Lower mantle ........................................................................................................................ 312 Core....................................................................................................................................... 314 PROBLEMS AND OPPORTUNITIES .......................................................................................316 CONCLUSIONS ..........................................................................................................................322 REFERENCES.............................................................................................................................323

11

Modeling Deformation of Polycrystalline Rocks Paul R. Dawson

MODELING PRELIMINARIES .................................................................................................331 Length scales..........................................................................................................................331 General comments .................................................................................................................332 Small-scale simulations .........................................................................................................332 Large-scale simulations .........................................................................................................333 SINGLE CRYSTAL CONSTITUTIVE BEHAVIOR .................................................................334 Slip systems ...........................................................................................................................334 Crystal kinematics..................................................................................................................334 Crystal compliance and stiffness ...........................................................................................335 POLYCRYSTAL CONSTITUTIVE EQUATIONS....................................................................336 Orientational averages ...........................................................................................................336 Linking crystal responses to continuum scale motion ...........................................................337 SMALL-SCALE HYBRID ELEMENT FORMULATION.........................................................337 LARGE-SCALE VELOCITY-PRESSURE FORMULATION...................................................339 HALITE TEXTURE EVOLUTION: A SMALL-SCALE APPLICATION................................340 Generalities ............................................................................................................................340 Simulation specifics ...............................................................................................................342 Simulation results ..................................................................................................................342 MANTLE CONVECTION — A LARGE-SCALE APPLICATION ..........................................343 Generalities ............................................................................................................................343 Simulation specifics ...............................................................................................................345 Simulation results ..................................................................................................................346 SUMMARY .................................................................................................................................350 ACKNOWLEDGMENTS............................................................................................................350 REFERENCES.............................................................................................................................350

12

Seismic Anisotropy and Global Geodynamics Jean-Paul Montagner, Laurent Guillot

INTRODUCTION........................................................................................................................353 CAUSES OF SEISMIC ANISOTROPY FROM MICROSCOPIC TO LARGE SCALE ...........354 Shape Preferred Orientation (S.P.O.).....................................................................................354 Lattice Preferred Orientation (L.P.O.) ...................................................................................355 EFFECT OF ANISOTROPY ON SEISMIC WAVES.................................................................357 Body waves............................................................................................................................358 Surface waves ........................................................................................................................359 Comparison between surface wave anisotropy and SKS splitting data .................................363 ANISOTROPY IN THE DIFFERENT LAYERS OF THE EARTH AND THEIR GEODYNAMIC APPLICATIONS ....................................................................365 Reference 1-D Earth models ..................................................................................................366 Evidence of anisotropy in the upper 410 km of the mantle ...................................................367 Oceanic plates ........................................................................................................................367 Continents ..............................................................................................................................370 Anisotropy in the transition zone ...........................................................................................372 Anisotropy in the D′′-layer ....................................................................................................372 NUMERICAL MODELING AND BOUNDARY LAYERS ......................................................373 ACKNOWLEDGMENTS............................................................................................................377 APPENDIX A. BASIC THEORY OF WAVE PROPAGATION IN ANISOTROPIC MEDIA...........................................................................................................377 APPENDIX B. TENSORS AND MATRICES MANIPULATIONS .........................................378 REFERENCES.............................................................................................................................380

13

Theoretical Analysis of Shear Localization in the Lithosphere David Bercovici, Shun-ichiro Karato

INTRODUCTION........................................................................................................................387 THEORETICAL PRELIMINARIES ...........................................................................................388 SHEAR LOCALIZING FEEDBACK MECHANISMS ..............................................................389 Thermal feedback with decay-loss healing............................................................................390 Simple-damage feedback with decay-loss healing ................................................................396 Grain-size feedback ...............................................................................................................399 Thermal and simple-damage feedbacks with diffusive-loss healing .....................................405 OTHER CONSIDERATIONS .....................................................................................................411 Two-dimensional examples ...................................................................................................411 More sophisticated damage theories ......................................................................................413 SUMMARY AND CONCLUSIONS...........................................................................................413 REFERENCES.............................................................................................................................418

1

New Developments in Deformation Studies: High-Strain Deformation Stephen J. Mackwell Bayerisches Geoinstitut Universität Bayreuth 95440 Bayreuth, Germany

Mervyn S. Paterson Research School of Earth Sciences Australian National University Canberra, Australia

INTRODUCTION Plastic deformation of rocks in the Earth is most strikingly illustrated in exposures of shear zones. Very large strains (shear strains of γ = 10 or more) may be developed in these zones. They appear to be ubiquitous within the continental and oceanic crust and may provide the dominant locus for deformation in these regions and, potentially, also in the lithospheric upper mantle. It is therefore of the greatest interest to obtain information about the deformation properties of rock at large strains In the past, various approaches have been made to achieving in the laboratory the types of deformation seen in natural rocks. In some cases it has been possible to achieve plastic deformation in single crystals and fine-grained aggregates of minerals at atmospheric pressure at sufficiently high temperatures and a number of such experiments have been carried out (Mackwell et al. 1990, Dimanov et al. 1998). However, it is generally necessary to conduct experiments under conditions of high confining pressure as well as high temperature. The earliest experiments of this nature were those of Adams and Nicholson (1901) and von Kármán (1911), followed by the pioneering work of Bridgman (1949) and Griggs (1936, and many subsequent papers). These studies were carried out using fluid pressure media. Although Otto Mügge carried out some deformation experiments on single crystals of minerals in solid confining media already in the late 19th to early 20th century, the extensive use of solid media for high-pressure experiments was not introduced until the development of the Griggs solid medium apparatus in the 1960's (e.g., Griggs et al. 1966, Griggs 1967). Since the 1960s, extensive studies have been carried out in both gas-medium and solid-medium apparatus. The apparatus and techniques have been reviewed in several publications (Griggs et al. 1960, Paterson 1964, 1970, 1978, 1990, Tullis and Tullis 1986, Kohlstedt et al. 1995). These studies have concentrated on axisymmetric triaxial deformation (the so-called “triaxial test”). However, the total amount of deformation in such tests is limited by several considerations discussed below. Also, axisymmetric tests do not give information about rotational types of strain history such as occur in shear zones. Therefore attention has turned to shear tests in which both large strains and rotational strain histories can be studied. This paper deals with the various experimental approaches to achieving large deformations. EXPERIMENTAL METHODS FOR LARGE STRAINS Experimental studies of the mechanical behavior and textural evolution of rocks to high strains necessitate apparatus that permits large changes in shape or configuration of 1529-6466/00/0051-0001$05.00

2

Mackwell & Paterson

samples while maintaining stable control for the purpose of measuring the deformation and the deforming forces. There are two types of homogeneous deformation path that can be studied, which can be distinguished as, respectively, coaxial and rotational. Either type of deformation path may be followed in attempting to reach large deformations. The properties and potentialities of these types of deformation for large strains are as follows: Coaxial deformation The principal axes of the strain increments the finite coaxial deformation. This is illustrated shortening deformation. The “true” measure of integrated strain based on current length l , that is,

remain parallel to each other during in Figure 1a for the case of simple the finite strain in this case is the l e = − ∫ dl l = − ln 1 l where l1 and l0 are 0

the final and initial lengths, respectively (shortening strain positive). Practical examples of coaxial deformation tests include the triaxial compression and extension tests.

Figure 1. Schematic of the evolution of the principal strain axes of samples deformed in (a) axial compression, (b) simple shear and (c) using the split-cylinder shear configuration.

If the tests are carried out to large strains in compression, the increase in crosssectional area tends to lead to the specimen overlapping the compression platens unless it is re-machined to smaller diameter at stages during the deformation. Further, the associated decrease in length leads to an unfavorable length-to-diameter ratio, with greater distortion of the stress field due to end-friction effects. In extension, the uniform strain is limited in magnitude by necking instability of the specimen (Hart 1967). Relatively large strains may develop within the neck (Griggs et al. 1960) but the stresses and strains are now inhomogeneous and a sophisticated analysis is required to deduce stress-strain properties in this case (Bridgman 1952). Rutter (1998) has carried out extension tests to obtain information about the behavior of marble in large coaxial deformation, with natural strains up to about γ = 2.4. In either case, there are practical limitations in accommodating the displacements involved in reaching large strains. In compression, the increase in diameter of the specimen may lead to interference with the bore of the furnace in the deformation apparatus. In extension the maximum displacement may be limited by the range of movement in the loading system or by the length of the region of uniform temperature in the furnace. Rotational deformation The principal strain axes rotate relative to external coordinates during the finite

High-Strain Deformation

3

deformation, as illustrated in Figure 1b for simple shearing deformation. Simple shear is an idealization approximated in greater or less degree in several types of experimental shearing arrangements under high pressure. These include the diagonally split-cylinder test and the torsion test. In the diagonally split-cylinder shear test, a non-deformable cylinder is cut at an acute angleθ, usually 45°, to the cylindrical axis and the deformable specimen inserted as a layer within the cut (Fig. 1c). The cylinder is then loaded axially, giving rise to a shear loading on the deformable layer. This arrangement can be conveniently inserted in a triaxial testing rig with a jacket enclosing the specimen assembly. However, it should be noted that the stress distribution is not that of a simple shear test because there is an additional normal stress acting on the layer. Moreover, in addition to the axial force F applied by the testing machine, there will also be a lateral force P acting on the layer due to constraints brought into play by the lateral component of the shearing displacement. Thus, as shown in the Appendix I, the shear stress τ parallel to the layer will be

τ=

2 F ⎛P 1 + ⎝ ⎞⎠ cosθ ′ sinθ A F

where

θ′ =θ +

π 2

− arctan

F , P

which will be accompanied by a normal stress σ n on the layer of

σn =

F ⎛P 2 1 + ⎝ ⎞⎠ sin θ ′ sin θ . A F

The presence of the lateral force component P is often not acknowledged but it may be quite important, especially in situations such as in a solid-medium machine where the loading piston is short and the medium also provides a resistance to lateral displacement. The actual stresses are difficult to calculate in such a case, but an estimate can be made of the lateral constraints in a fluid-medium machine where it corresponds to the bending of the loading pistons (see Appendix I). Thus additional stress components of tens of MPa can arise from lateral constraints. This effect will be most serious for weak specimens. The other important feature of the split-cylinder shear test is the presence of the substantial normal stress component on the layer in addition to the shear stress. The normal stress tends to lead to thinning of the specimen layer and its lateral extrusion. The advantage of the split-cylinder shear test is that it is often a convenient way of achieving substantial shear strains in thin layers of material, permitting the study of the evolution of microstructures to moderately large strains. However, it must be borne in mind that the stress conditions are not simple to analyse and that the deformation may include substantial flattening strains as well as the shear strains. The torsion test is, in effect, another type of shear test. It has proved to be the most effective way of achieving very large strains. Moreover, the non-coaxial deformation path closely simulates that in geological shear zones. The deformation conditions at any point in a cylindrical torsion specimen correspond to simple shear and are not complicated by the presence of a normal stress component as in the split-cylinder test, provided no axial force is applied to the cylinder. The plane of shearing is normal to the axis of the cylinder. Practical aspects of the torsion test and its application are given in Appendix II. The analysis of the torsion test has been set out by Paterson and Olgaard (2000). In torsion of a cylinder, the shear strain varies linearly with the radius, being a maximum at the outer surface of the cylinder. If the angle of twist is θ in a cylinder of diameter d and length l , then the maximum shear strain γ at the surface of the cylinder is given by

4

Mackwell & Paterson

γ=

dθ 2l

However, the stress distribution is not linear with radius and depends on the nature of the stress-strain relationship. If the form of the stress-strain relationship can be assumed, for example, if it can be expected to be of power-law form, the torque-twist observation can be inverted to obtain the stress-strain relationship. Thus if a power law with stress exponent n is assumed, the shear stress τ at the surface of the specimen is given by 1 ⎛ 4M ⎝ 3 + ⎞⎠ n τ= 3 πd where M is the measured torque. Since the stress increases with radius more rapidly than linearly for n > 1, and the contribution of each radial element is proportional to the radius, most of the torque derives from the stress in the outer layers of the specimen. Thus the circumstance that there is a radial gradient of stress and strain in the specimen is generally of relatively minor importance, and Paterson and Olgaard (2000) argue that there is little point in attempting to carry out experiments with hollow specimens in order to achieve more nearly homogeneous deformation, especially in view of the experimental difficulties of jacketing specimens inside and outside. Microstructural observations are most appropriately made on thin sections cut tangentially to the outer surface of the specimen. In this case, the section represents the plane of simple shearing. Texture observations can also be made in such a section. LARGE-STRAIN BEHAVIOR IN ROCKS — GENERAL CONSIDERATIONS

Considerations of large-strain behavior in rocks call for a broader view of time/strain scales than has been usual for small-strain studies. In particular, new perspectives on “steady state” are needed. In constant strain-rate axial deformation tests, it is common for the stress-strain curve to become more or less flat, that is, strain-independent, after a true strain of between 0.1 and 0.3. The behavior is said to have reached a “steady state”, with the implication that the behavior has become independent of strain and that the structure or state of the material is not longer changing significantly. However, observations on specimens at this stage often reveal that a partial degree of recrystallization or some change in the grain size has occurred and that there is the potential for further change. Thus factors that influence the level of the flow stress have become nearly constant over the relatively short strain intervals of order 0.1 but in fact slower or longer-term changes are still in the process of occurring. Torsion tests on rocks so far have revealed that evolutionary changes such as recrystallization tend to require strains of at least γ = 5 or more to come to completion and that the stress-strain curve also shows non-steady-state trends over this range (Fig. 2). It is also possible that grain growth can proceed over even longer strain intervals. Since it has become clear from large strain studies that apparent “steady state” behavior after relatively small strains does not generally persist over larger ranges of strain, interest has shifted more to the evolutionary aspects of deformation properties and structural change. The quantitative description of the mechanical behavior is no longer simple, as embodied in the steady-state power law under the assumption of isotropic behavior, and it is difficult to establish simple expressions that describe stress-strain curves such as those now commonly obtained from large-strain torsion tests (e.g., Pieri et al. 2001a, Mackwell and Rubie 2000). Also a treatment that takes into account the developing anisotropy has not yet been elaborated. An alternative approach is to use an

High-Strain Deformation

5

expression that is valid over a limited part of the stress-strain curve and determine the variation in the parameters over the full strain range to give a full description. Such an expression is

γ = Aγ − pτ n exp(− Q / RT )

(1a)

or −

1 n

p n

1

τ = A γ γ n exp(Q / nRT )

(1b)

where τ is the shear stress, γ the shear strain, and T the temperature (R is the gas constant). In theoretical continuum mechanics it is sometimes preferred to write the temperature term as Tν rather than in the Arrhenius form above (e.g., Fressengeas and Q Molinari 1987); in this case, ν = − . nRT0

Figure 2. Rheological, textural and microstructural evolution of Carrara marble deformed to high shear strain in torsion. Shear stress τ, stress exponent n, texture index J and area fraction of recrystallized grains are given as a function of shear strain γ. The strain intervals for the various types of experiments are given at the bottom. [Used by permission of the editor of the Journal of Structural Geology, from Pieri et al. (2001), Fig. 17, p. 1411.]

The quantities A, p, n and Q are now parameters whose values gradually change or evolve in the course of the deformation but which can be taken as constants over limited ranges of strain. Parameters n and Q give measures, respectively, of the stress sensitivity and temperature sensitivity of the strain rate, and the ratio p/n is a measure of the strain hardening. Thus ∂ ln γ n= at constant γ and T , ∂ ln τ ∂ ln γ ∂ ln γ Q = −R = RT 2 at constant γ and τ ∂ (1 / T ) ∂T or

6

Mackwell & Paterson Q = nR

and

∂ ln τ 2 ∂ ln τ = −nRT ∂ (1/ T) dT

at constant γ and γ ,

p=n

∂ ln τ ∂ ln γ

p=−

∂ ln γ at constant τ and T (creep test) d ln γ

at constant γ and T (stress-strain test)

or

The last expressions give directly a measure of the strain hardening in terms of the slope of the stress-strain curve or creep curve, so dτ p τ = at constant γ and T (stress-strain test) dγ nγ or dγ γ = −p at constant τ and T (creep test). γ dγ Parameter A is a related to the absolute level of the stress-strain curve, and it can also evolve with strain. In special cases, A can be expanded to show an explicit variability of the flow behavior with particular factors such as initial grain size or other internal variables, oxygen or water fugacity, etc. If the parameters in (1) are constant and p = 0, we have the “steady state” flow law that is often quoted in geodynamical discussions. It is now evident that the steady state discerned at strains of order 0.2 is commonly ephemeral but there may be an approximation to steady state again at large strains, with γ > 5 or 10. LARGE-STRAIN BEHAVIOR IN ROCKS — CASE STUDIES

There have been a number of studies in recent years of the high-strain behavior of rocks using the techniques described in the previous sections. In some of these studies, measurements have been made on the same rock type using complementary techniques. Many of these measurements have been performed relatively recently, notably due to technological developments, such as the utilization of the torsional configuration in the gas-medium apparatus. In the following, we will review the current state of the field in high-strain experimental studies of rocks, focussing on monomineralic crustal and mantle rocks and then briefly summarizing research on two-phase and polyphase systems. Monomineralic crustal rocks

Perhaps the most studied rock in terms of deformation behavior, Carrara marble has been investigated using a range of techniques, under compressional, extensional, splitcylinder and torsional configurations. The axial compression studies have provided a comprehensive understanding of low-strain deformation mechanisms, notably dislocation creep (e.g., Rutter 1974, Schmid et al. 1980). Under extensional deformation conditions to high strains, Rutter (1998) performed experiments at 973 and 1073 K, utilizing the necking instability to attain maximum shear strains of around γ = 2.4. He reported significant dynamic recrystallization at high strains. The observed weakening with progressive strain may have resulted from purely geometric constraints due to necking at constant displacement rate. A further study that made use of the split-cylinder configuration (Schmid et al. 1987), which allowed constant strain rates, also reported

High-Strain Deformation

7

significant grain-size reduction due to dynamic recrystallization but only a modest reduction in shear strength to a shear strain of around γ = 3. In neither case was a change in deformation mechanism from dislocation creep to diffusional creep inferred despite the significant grain refinement.

Figure 3. Torque versus shear strain for all constant strain rate tests on Carrara marble at 1000 and 1200 K. The torque is converted into shear stress on the scale to the right assuming 15 mm (inset 10 mm) diameter samples and stress exponent of n = 10. [Used by permission of the editor of the Journal of Structural Geology, from Pieri et al. (2001), Fig. 6, p. 1399.]

More recently, Pieri et al. (2001a,b) used the torsional configuration to study highstrain deformation of Carrara marble at 1000 - 1200 K to shear strains as high as γ = 11. After an initial peak at a shear strain of around γ = 1, the strength decreased to a steady level at a shear strain of around γ = 5 (Fig. 3). The low-strain microstructures showed undulose extinction, deformation bands and the development of subgrain boundaries within the grains, as well as some bulging of grain boundaries (Fig. 4). A strong shape preferred fabric developed with increasing strain, but was destroyed as the samples became fully recrystallized by a strain of between γ = 5 and 11, depending on temperature. The grain-size reduction from around 150 to 10 µm occurred predominantly due to nucleation and growth of strain-free grains on grain boundaries. The lattice preferred orientation that developed at low strains, which was consistent with the slip systems identified in axial compression experiments, was progressively replaced by a recrystallization texture. Experiments on fine-grained Solnhofen limestone show behavior that is quite distinct from that the coarser grained marble. Low-strain axial compression experiments show several deformation fields. At higher-stress, lower-temperature conditions, deformation is dominated by dislocation creep, while at lower stresses behavior is consistent with a mechanism dominated by grain-boundary sliding accommodated by diffusion along grain boundaries (Schmid et al. 1977). Schmid et al. interpreted this latter behavior as “superplastic” creep. They observed no net grain growth (due presumably to the presence of impurity phases) and no associated changes in deformation mechanism during the experiments. Studies of the texture development in such samples (e.g., Rutter et al. 1994) showed strong development of a lattice preferred orientation in the dislocation creep field, but only weak development in the “superplastic” field. Using the split-cylinder configuration, Schmid et al. (1987) extended the work to shear strains of γ ≅ 1, while Casey et al. (1998) used the torsion design to reach shear strains of up

8

Mackwell & Paterson

Figure 4. Microphotographs in cross-polarized light of Carrara marble samples deformed in torsion at 1000 K to shear strain of γ = 1 (a,b), γ = 2 (c,d), γ = 5 (e,f) and γ = 11 (g,h). Overviews at low magnifications (at left) and details at higher magnification (at right). Shear zone boundary is oriented parallel to the long edge of the image and the sense of shear is dextral. The arrows in (b) point to a small relict grain and to bulging of the grain boundary, indicating that the horizontal grain boundary is migrating downwards. [Used by permission of the editor of the Journal of Structural Geology, from Pieri et al. (2001), Fig. 8, p. 1402.]

High-Strain Deformation

9

to γ = 12. Within the region near the transition from dislocation to “superplastic” creep, little further textural or mechanical evolution was observed beyond that seen in the lowstrain axial experiments, as illustrated in Figure 5. The presence of a lattice preferred orientation and near-equant grains argues for deformation by both dislocation creep and grain boundary sliding. At high strains, both textures and mechanical behavior appeared to be near steady state. Another material that has been the focus of many deformation studies is quartzite. Deformation to small strains in compression has been performed on a range of quartzites with large variability in grain size, impurity phase and water contents. Deformation under dry conditions yields highly variable rheologies characterized by mostly brittle mechanisms. Under wet high-pressure conditions, quartzites can deform by fully plastic mechanisms, but the measured rheologies are highly variable in terms of the calculated constitutive parameters (e.g., Kronenberg and Tullis 1984, Luan and Paterson 1992, Gleason and Tullis, 1995). High-strain deformation experiments on untreated (and hence wet) samples of several quartzites (mostly flint) in torsion by Schmocker et al. (2002) reveal deformation by granular flow (a brittle to semi-brittle deformation mechanism), with formation of bands of localized deformation, which appear to form in regions of excess fluid pressure, and development of no significant texture or shape fabric. Anhydrite is the only material studied to date that shows coupled textural evolution and changes in deformation mechanism (Stretton and Olgaard 1997). As with the marble, recrystallization occurs predominantly by nucleation and growth of strain-free grains on grain boundaries. The high-strain experiments have been performed in torsion. Bulging of grain boundaries begins at relatively low shear strains of around γ = 0.5 and recystallization is essentially completed by γ = 3.7. A parallel and relatively abrupt strength reduction occurs as the deformation mechanism changes from dislocation to diffusional creep at a shear strain of about γ = 1.5 (Fig. 5). Subsequently samples appear to reach a near-steady-state microstructure and rheology. Deformation textures develop quite rapidly at lower shear strains but weaken with progressive deformation and recrystallization (Heidelbach et al. 2001). After recrystallization, there is no evidence of significant grain growth to a shear strain of γ = 20, stabilizing deformation in the grain boundary diffusional creep field. The reason for the inhibition of grain growth in the diffusion creep field despite initial rapid growth of recrystallized grains in the dislocation creep field is not clear. In this case, pinning by impurity phases seems unlikely. For this material progressive recrystallization with increasing strain in torsion may result in a sample with a rim deforming by diffusional creep surrounding an interior deforming by dislocation creep, complicating interpretation of the stress distribution within the sample. However, there is no clear signature of such a complex deformation geometry in the mechanical behavior, which shows a stress exponent of near n = 1 at higher strains (Stretton and Olgaard 1997). Monomineralic mantle rocks

A significant body of both low- and high-strain deformation data has been collected for magnesiowüstite, the second most abundant phase in the Earth’s lower mantle. While important as a component of aggregate flow, it is not clear that single-phase flow has a direct relevance to convection in this region of the Earth. Stretton et al. (2001) have performed axial compression experiments on (Mg0.8Fe0.2O) to determine the low-strain mechanical behavior. The material deforms in the dislocation creep regime with a stress exponent of near n = 4, with the development of a weak lattice preferred orientation, consistent with the dominant slip systems for rock-salt-structure materials. High-strain experiments have been performed in both torsion (Heidelbach et al. 2002) and split-sphere (Yamazaki and Karato 2002) configurations. The torsion experiments on

10

Mackwell & Paterson

Figure 5. Shear stress versus shear strain for tests on polycrystalline samples of magnesiowüstite (Heidelbach et al. 2002), olivine (Bystricky et al. 2000), Solnhofen limestone (Casey et al. 1998) and anhydrite (Stretton and Olgaard 1997) at the temperatures and shear strain rates shown on the right. The three small images are microstructure maps derived from SEM electron backscatter diffraction investigations of tangential sections of magnesiowüstite samples deformed to shear strains of γ = 0, 2.3 and 13.

(Mg0.8Fe0.2O) show deformation by dislocation creep (the low-strain behavior is largely consistent with the axial experiments) with the development of a strong shape preferred fabric and initiation of dynamic recrystallization by progressive subgrain rotation (Fig. 6). Full recrystallization occurs by a shear strain of about γ = 6 with loss of the shape preferred fabric. The initial lattice preferred orientation, which is consistent with the dislocation slip systems expected for the rock-salt structure, weakens with progressive strain and is overprinted by a recrystallization texture. No change in deformation mechanism has been observed and, after an initial peak stress (in some cases) the strength of the samples remains essentially constant to shear strains as high as γ = 16 (Fig. 5). While the split-sphere experiments of Yamazaki and Karato (2002) on (Mg0.75Fe0.25O) were not able to access the full range of conditions sampled in the torsion experiments (maximum γ = 8), the overall conclusions are essentially the same. High-strain deformation studies have also been performed on upper mantle rocks. Of particular note, dry olivine aggregates have been deformed to high strains in torsion (Bystricky et al. 2000) and using the split-cylinder configuration (Zhang and Karato 1995, Zhang et al. 2000). The low-strain behavior in these tests is consistent with the

High-Strain Deformation

Figure 6. Microstructure maps derived from SEM electron backscatter diffraction investigations of tangential sections of magnesiowüstite samples, with a map size 1 × 1 mm2, and step width 5 µm; (a) γ = 1.1, (b) γ = 3.1, (c) γ = 15.5, (d) γ = 2.3, (e) γ = 5.8, (f) γ = 13; (a-c) T = 1300 K, (d-f) T = 1400 K; grain boundaries were drawn in black for misorientation angles ω > 10°. [Used by permission of the authors, from Heidelbach et al. (2002), Fig. 6.]

11

12

Mackwell & Paterson

large body of data on rocks of similar compositions from axial compression tests performed in the dislocation creep regime (e.g., Chopra and Paterson 1984, Karato et al. 1986). At larger strains recrystallization by progressive subgrain rotation is apparent resulting in a net reduction in grain size. No change in deformation mechanism has been observed. Although some bulging of grain boundaries is apparent in the optical thin sections, subgrain rotation seems to be the dominant recrystallization mechanism. As with the marble and magnesiowüstite, a strong shape preferred fabric develops at lower strains but is destroyed by the recrystallization. The lattice preferred orientations measured for the high-strain samples are consistent with the dislocation slip systems observed in low-strain experiments. Interestingly, although both Bystricky et al., and Zhang et al. reported textures with the [100] axis parallel to the shear direction, Zhang et al. noted a (010) plane parallel to the shear plane while Bystricky et al. showed girdles of [010] and [001] normal to the shear direction. Unlike the magnesiowüstite, the aggregate strength decreases steadily with increasing strain to the maximum reported shear strain of around γ = 5 (Fig. 5). Recent experiments on olivine aggregates under hydrous conditions by Jung and Karato (2001b) using the split cylinder configuration yield lattice preferred orientations consistent with a change from the dominant a-slip (dominance of [100] Burgers vector dislocations) to c-slip (dominance of [001] dislocations) with increasing water content. While the cause of the changing textures is debated (Kaminski 2002), such a difference in the deformation textures for wet and dry regions would have important ramifications in the interpretation of seismic anisotropy observations for the lithospheric and asthenospheric mantle. Tullis and Yund (1999) and Stünitz et al. (2001) have performed experiments on plagioclase aggregates under predried and water-added conditions to shear strains of around γ = 2 using the split-cylinder configuration in the solid-medium apparatus. They observed that the predried plagioclase stayed within the dislocation creep regime but showed some weakening with increasing strain, while the wet plagioclase deformed in the diffusional creep field. Transmission electron microscopy images of the predried samples showed remanent high-dislocation-density grains surrounded by small polygonal recrystallized grains. A preliminary study of high-strain deformation of clinopyroxenite to relatively high strains by Bystricky and Mackwell (2001) yielded the same general observations as for olivine, with development of an initial shape-preferred fabric, subgrain rotation recrystallization and maintenance of dislocation creep as the dominant deformation mechanism. The observed lattice preferred orientations are also consistent with the dislocation slip systems operative in previous low-strain axial deformation experiments. Polyphase systems

Obviously the Earth is not comprized of significant regions where the rock is monomineralic and an important challenge for the future is to characterize the deformation behavior of multi-phase systems, both to low and high strains. In particular, we aim to mimic processes occurring in plastic shear zones (regions where deformation occurs by plastic deformation that is relatively localized spatially). Such zones are often the downward continuation of shallower brittle fault systems. Of special interest, we would like to see how the individual minerals contribute to the deformation and whether the deformation leads to phase separation into interlayered structures where deformation is concentrated in the layers comprized of the weaker minerals. Several recent studies have begun to address these questions. Deformation experiments in torsion have been performed on mixtures of quartz and

High-Strain Deformation

13

calcite with variable modal abundances by Rybacki et al. (2002). In these experiments, the quartz essentially acts as a non-deforming component in a plastic calcite matrix and is observed to exert a strong effect on aggregate strength. To a large extent the behavior mimics the high-strain deformation of marble, but strengthened by the presence of a second harder phase. Rybacki et al. also observe a strong temperature effect on the extent of recrystallization and strain softening behavior even to high strains. Experiments have also been performed to high strains in 2-phase systems where the individual phases are not so different in strength. Barnhoorn et al. (2002) chose a range of modal compositions between pure anhydrite and pure calcite and investigated deformation under conditions where the two phases have similar strengths. For a shear strain of about γ = 1, the pure anhydrite samples show significant weakening relative to pure calcite, with intermediate behavior for mixed aggregates. With increasing strain, the steady-state stress values for all compositions converge, but deformation becomes localized, with phase separation, seemingly mimicking observations of natural plastic shear zones. While beyond the scope of the present review, there is some interesting work currently investigating the effect of high shear strains on the texturing of melt within partially molten systems (e.g., Zimmerman et al. 1999, Bruhn et al. 2000, Holtzman et al. 2002), with implications for such diverse processes as core formation and melt segregation at mid-ocean ridges. SUMMARY

While numerous experiments have been performed to high strains on a variety of materials using the various technologies described in this paper, this is still a relatively new field of study and much of the work reported above is research in progress. In the near future, we will certainly see major advances, particularly using the torsional configuration of the gas-medium apparatus. There are, however, a number of interesting observations that can be made based on the experimental studies to date on high-strain deformation of earth materials: • In almost all materials to date, dynamic recrystallization resulted in a reduction of grain size, although only in the case of anhydrite did that result in a change in deformation mechanism. Recystallization occurred by subgrain rotation, by nucleation and growth of new grains on grain boundaries, or a combination of both. Complete recrystallization required shear strains of at least γ = 3, which is considerably higher than can be attained in axial compression experiments. Thus, textural and mechanical steady state can only be attained during high-strain experiments. • Strain softening was observed even to relatively high shear strains in a number of cases, presumably reflecting progressive alignment of favorable slip systems or incomplete recrystallization. • In cases where shear strains higher than γ = 10 have been attained for hightemperature deformation in the dislocation creep field, initial textures reflecting the active dislocation slip systems evolve towards a recrystallization texture. • The role of water in terms of textural and mechanical behavior has been relatively little studied to date. However, the results for olivine (Jung and Karato 2001a,b) and plagioclase (Tullis and Yund 1999, Stünitz et al. 2001) suggest that it may play a major role, with important implications for the behavior of the Earth’s interior.

14

Mackwell & Paterson APPENDIX I: STRESS ANALYSIS FOR DIAGONALLY SPLIT-CYLINDER SHEAR TEST

The specimen layer in the diagonally split-cylinder test is subjected to a force F parallel to the cylinder axis, apart from the hydrostatic loading from the confining pressure. If the cylinder cut is at an angle θ to the cylinder axis, then the layer can be regarded as being initially subject to a shear force component F cosθ parallel to the layer and a normal force component F sinθ normal to the layer. If the cylinder is of crosssectional area A , the layer is initially subject to a shear stress component (F/A) cosθ sinθ and a normal stress component (F/A) sin2θ. However, this stress state strictly only applies at the initial stage of loading. As soon as some shearing displacement takes place, there is also a lateral component of displacement, normal to the cylinder axis, which will result in a reaction force from the constraints of the testing machine that are intended to ensure primarily axial displacement. As an illustration, suppose that the split cylinder is loaded by two pistons of diameter D and length L , and that a shearing displacement s has occurred across the specimen layer. This displacement has a lateral component s sinθ, requiring a lateral displacement (s/2) sinθ of the end of each piston. Suppose that the loading pistons are rigidly constrained at their outer ends and that the lateral deflection of the inner ends is accommodated entirely by elastic bending of the pistons. Since the deflection of the end 3 of a cantilever beam is given by PL /3EI (Gere and Timoshenko 1984, p. 369), where P is the lateral force, E the Young's modulus, and I = πD4/64, the moment of inertia of the beam cross-section, we then have P=

3π sinθ ED 4 s 128 L3

For example, if D = 15 mm, E = 200 GPa, L = 100 mm, and θ = 45˚, then P = 527 ⎛⎜⎝

s ⎞ ⎟ mm ⎠

N,

or 527 N if s = 1 mm. If the split cylinder were of 10 mm diameter, this lateral force would thus introduce an additional lateral component of normal stress of about 7 MPa when the shearing displacement is 1 mm. The resultant force R on the specimen layer is therefore 2 ⎛P R = F 2 + P2 = F 1+ ⎝ ⎞⎠ F

which is inclined at an angle θ ′ to the specimen layer, where

θ′ =θ +

π 2

− arctan

F P

The resultant shear stress τ parallel to the layer is then

τ=

2 F ⎛P 1 + ⎝ ⎞⎠ cosθ ′ sinθ A F

and the normal stress σ n acting on the layer is

σn =

2 F ⎛P 1 + ⎝ ⎞⎠ sin θ ′ sin θ . A F

The principal stress components in the plane of the cylinder axis and the normal to the

High-Strain Deformation

15

specimen layer, in addition to the hydrostatic stress, are ⎛ σn ⎞ 2 σ1,2 = ± ⎝ ⎠ +τ2 2 2

σn

and their axes are inclined to the specimen layer at angles α and π/2 – α such that

σn

cos2α =

2

2

⎛ σn ⎞ + τ2 ⎝ 2 ⎠

and sin2α =

τ 2

⎛σn ⎞ +τ2 ⎝ 2 ⎠

(Gere and Timoshenko 1984, p. 288) APPENDIX II: THE TORSION TEST A discussion of the torsion test as applied to experimental rock deformation has been given by Paterson and Olgaard (2000). References are given there to applications in materials science and to earlier applications in the testing of rocks, ice, and rock gouges. Notable was the pioneering work of Handin et al. (1960, 1967). The particular test configuration described by Paterson and Olgaard (2000) is an add-on module for the Paterson gas-medium HPT testing machine (Paterson 1990); see Figure A1. The combination with an axial deformation facility gives considerable flexibility of operation, including the ability to check length changes in specimens during torsion and even to apply axial force in combination with a torque so as to vary all three principal stresses independently (Handin et al. 1967). By using an internal furnace, temperatures of up to 1600 K can be reached at confining pressures of up to 500 MPa. A combined internal load cell permits the measurement of axial force and torque simultaneously, while avoiding uncertainty from friction corrections. The most convenient way of gripping the specimen is to utilize the friction between the driving piston and the specimen, arising from the confining pressure, to transmit the torque to the specimen. However, for very strong specimens, inadequate frictional force may lead to slipping at this interface. In such cases, it may be necessary to resort to “dogbone” shaped configurations to increase the available torque. In any case, it is important to vent the inside of the jacketed assembly in order not to reduce the friction by the presence of a pore fluid pressure. Conversely, this consideration markedly limits the scope for combining pore fluid studies with torsion testing. Jacketing itself may present problems with very large strain tests because the jacket has also to be capable of undergoing the same strains as the specimen without disintegrating. Thus it has been observed that iron (low-carbon steel) is capable of very large strains at moderate temperatures, up to around 1000 K, but at higher temperatures it seems to develop excessive grain growth and to disintegrate by parting at the grain boundaries at strains of γ = 5 to 10. The selection of appropriate jacket material is part of the preparation for high-strain experiments. An advantageous aspect of torsion testing is that end effects associated with friction at the contact with the driving pistons are much less than in axial deformation tests where friction gives rise to constraints on the change in diameter. Thus, if need be, very short specimens can be used. Another aspect of torsion testing is that shear localization with respect to the axial distribution of deformation can develop with little constraint if unstable behavior arises in the outer layers of the specimen. However, theoretical considerations (Fressengeas and Molinari 1987) suggest that localization is unlikely to develop in torsion tests with even moderately large strain softening rates if the tests are conducted at constant twist rate, but may be more likely in creep tests at constant torque.

16

Mackwell & Paterson

Figure A1. Schematic diagram of high-pressure high-temperature testing machine with torsion actuator and specimen assembly. [Used by permission of the editor of the Journal of Structural Geology, from Paterson and Olgaard (2000), Fig. 5, p. 1346.]

The strain rates that can be used in torsion testing are limited by time constraints if large strains are to be achieved. Supposing that it were desired not to exceed about one day (say, 105 seconds) for a given experiment, then to achieve a strain of γ = 10 the strain rate would need to be not less than about 10-4 s-1. Thus the trend is to use higher strain rates in large-strain torsion tests than in axial deformation tests. However, the strain rate dependence of the flow stress of the material at any point in its evolution can be tested by strain-rate stepping to lower strain rates. The stress exponent n for a power-law fit can be obtained directly from the torque-twist measurements through n=

d ln θ d ln M

where θ is the twist rate and M is the torque. Similarly the apparent activation energy Q can be obtained from

High-Strain Deformation

Q = nR

17

d ln M d(1/ T )

(Paterson and Olgaard 2000). In order to compare rheological behavior in torsion and axial deformation, some assumption must be made concerning the nature of the constitutive relationship for general stress states. If the von Mises-type relationship for isotropic behavior is assumed under which the flow behavior is the same for a given shear strain energy, then if the flow stress in a conventional triaxial is σ the flow stress τ in torsion for the same 1 “equivalent” strain rate is τ = σ . Alternatively, if the comparison is made at the same 3

nominal strain rate, we have τ =

1 1+n 2n

1 3

σ , varying from τ = σ to τ =

3

1 σ 3

as n increases

from 1 to infinity (Paterson and Olgaard 2000). Thus, to a rough approximation, the axial deformation flow stress (“differential stress”) is generally about double the shear flow stress in torsion. REFERENCES Adams FD, Nicolson JT (1901) An experimental investigation into the flow of marble. Phil Trans R Soc A195:363-401 Barnhoorn A, Bystricky M, Kunze K, Burlini L (2002) Localization of deformation in a two-phase rock: torsion experiments on calcite-anhydrite aggregates. EMPG IX J Conf Abstr 7:10 Bridgman PW (1949) The Physics of High Pressure (reprinted with Supplement), G. Bell and Sons, London, 445 p Bridgman PW (1952) Studies in Large Plastic Flow and Fracture. McGraw Hill, New York, 362 p Bruhn D, Groebner N, Kohlstedt DL (2000) An interconnected network of core-forming melts produced by shear deformation. Nature 403:883-886 Bystricky M, Kunze K, Burlini L, Burg J-P (1999) Rheology and texture of olivine aggregates deformed to high strains in torsion. EOS Trans AGU 80:F973. Bystricky M, Kunze K, Burlini L, Burg J-P (2000) High shear strain of olivine aggregates: rheological and seismic consequences. Science 290:1564-1567 Bystricky M, Mackwell S (2001) Creep of dry clinopyroxene aggregates. J Geophys Res 106: 13,443-13,454 Casey M, Kunze K, Olgaard DL (1998) Texture of Solnhofen limestone deformed to high strains in torsion. J Structural Geology 20:255-267 Chopra PN, Paterson MS (1984) The role of water in the deformation of dunite. J Geophys Res 89: 7861-7876 Dimanov A, Dresen G, Wirth R (1998) High-temperature creep of partially molten plagioclase aggregates. J Geophys Res 103:9651-9664 Fressengeas C, Molinari A (1987) Instability and localization of plastic flow in shear at high strain rates. J Mech Phys Solids 35:185-211 Gere JM, Timoshenko SP (1984) Mechanics of Materials. PWS Engineering, Boston, Massachusetts, 768 p Gleason GC, Tullis J (1995) A flow law for dislocation creep of quartz aggregates determined with the molten salt cell, Tectonophysics 247:1-23 Griggs DT (1936) Deformation of rocks under high confining pressures. J Geol 44:541-577 Griggs DT (1967) Hydrolytic weakening of quartz and other silicates. Geophys J R Astron Soc 14: 19-31 Griggs DT, Blacic JD, Christie JM, McLaren AC, Frank FC (1966) Hydrolytic weakening of quartz crystals. Science 152:674 Griggs DT, Turner FJ, Heard HC (1960) Deformation of rocks at 500° to 800°C. In Griggs D, Handin J (eds) Rock Deformation. Geol Soc Am Mem 79:39-104 Handin J, Heard HC, Magouirk JN (1967) Effects of the intermediate principal stress on the failure of limestone, dolomite, and glass at different temperatures and strain rates. J Geophys Res 72: 611-640

18

Mackwell & Paterson

Handin J, Higgs DV, O'Brien JK (1960) Torsion of Yule marble under confining pressure. In Griggs D, Handin J (eds) Rock Deformation, Geol. Soc. Am Mem 79, p 245-274 Hart EW (1967) The theory of the tensile test. Acta Metall 15:351-355 Heidelbach F, Stretton IC, Kunze K (2001) Texture development in polycrystalline anhydrite experimentally deformed in torsion. Intl J Earth Sciences 90:118-126 Heidelbach F, Stretton IC, Langenhorst F, Mackwell S (2002) Fabric evolution during high shear-strain deformation of magnesiowüstite (Mg0.8Fe0.2O). J Geophys Res (in press) Holtzman BK, Groebner NJ, Zimmerman ME, Ginsberg SB, Kohlstedt DL (2002) Deformation-driven melt segregation in partially molten rock. G3 (in press) Jung H, Karato S (2001a) Effect of water on the size of dynamically recrystallized grains in olivine. J Struct Geol 23:1337-1344 Jung H, Karato S (2001b) Water-induced fabric transition in olivine. Science 239:1460-1463 Kaminski E (2002) The influence of water on the development of lattice-preferred orientation in olivine aggregates. Geophys Res Lett 29(12):10.1029/2002GL014710 Karato S, Paterson MS, Fitz Gerald JD (1986) Rheology of synthetic olivine aggregates: influence of grain size and water. J Geophys Res 91:8151-8179 Kohlstedt DL, Evans B, Mackwell SJ (1995) Strength of the lithosphere: constraints imposed by laboratory experiments. J Geophys Res 100:17,587-17,602 Kronenberg AK, Tullis J (1984) Flow strengths of quartz aggregates: Grain size and pressure effects due to hydrolytic weakening. J Geophys Res 89:4281-4297 Luan F, Paterson MS (1992) Preparation and deformation of synthetic aggregates of quartz. J Geophys Res 97:301-320 Mackwell S, Rubie D (2000) Earth under strain. Science 290:1514-1515 Mackwell SJ, Kohlstedt DL, Durham WB (1990) High-resolution creep apparatus. In Duba AG, Durham WB, Handin JW, Wang HF (eds) The Brittle-Ductile Transition in Rocks. The Heard Volume. American Geophysical Union, Washington, DC, Geophys Monogr 56:235-238 Paterson MS (1964) Triaxial testing of materials at pressures up to 10,000 kg/sq cm (150,000 lb/sq in). J Institution Engin Australia 36:23-29 Paterson MS (1970) A high-pressure, high-temperature apparatus for rock deformation. Intl J Rock Mech Min Sci 7:517-526 Paterson MS (1978) Experimental Rock Deformation—The Brittle Field, Minerals and Rocks. SpringerVerlag, Berlin, 254 p Paterson MS (1990) Rock deformation experimentation. In Duba AG, Durham WB, Handin JW, Wang HF (eds) The Brittle-Ductile Transition in Rocks. The Heard Volume. American Geophysical Union, Washington, DC, Geophys Monogr 56:187-194 Paterson MS, Olgaard DL (2000) Rock deformation tests to large shear strains in torsion. J Struct Geol 22:1341-1358 Pieri M, Burlini L, Kunze K, Stretton I, Olgaard DL (2001a) Rheological and microstructural evolution of Carrara marble with high shear strain: results from high temperature torsion experiments. J Struct Geol 23:1393-1413 Pieri M, Kunze K, Burlini L, Stretton I, Olgaard DL, Burg J-P, Wenk H-R (2001b) Texture development of calcite by deformation and dynamic recrystallization at 1000 K during torsion experiments of marble to large strains. Tectonophysics 330:119-140 Rutter EH (1974) The influence of temperature, strain rate and interstitial water in the experimental deformation of calcite rocks. Tectonophysics 22:311-334 Rutter EH (1998) Use of extension testing to investigate the influence of finite strain on the rheological behavior of marble. J Struc Geol 20:243-254 Rutter EH, Casey M, Burlini L (1994) Preferred crystallographic orientation development in during the plastic and superplastic flow of calcite rocks. J Struct Geol 16:1431-1446 Rybacki E, Paterson MS, Wirth R, Dresen G (2002) Rheology of calcite-quartz aggregates deformed to large stain in torsion. J Geophys Res (submitted) Schmid SM, Boland JN, Paterson MS (1977) Superplastic flow in finegrained limestone. Tectonophysics 43:257-291 Schmid SM, Panozzo R, Bauer S (1987) Simple shear experiments on calcite rocks: rheology and microfabric. J Struct Geol 9:747-778 Schmid SM, Paterson MS, Boland JN (1980) High temperature flow and dynamic recrystallization in Carrara marble. Tectonophys 65:245-280 Schmocker M, Bystricky M, Kunze K, Burlini L, Stünitz H, Burg J-P (2002) Granular flow and Reidel band formation in water-rich quartz aggregates experimentally deformed in torsion. J Geophys Res (in review)

High-Strain Deformation

19

Stretton IC, Heidelbach F, Mackwell S, Langenhorst F (2001) Dislocation creep of magnesiowüstite (Mg0.8Fe0.2O). Earth Planet Sci Lett 194:229-240 Stretton I, Olgaard DL (1997) A transition in deformation mechanism through dynamic recrystallization evidence from high strain, high temperature torsion experiments. EOS, Trans Am Geophys Union 78:F723 Stünitz H, Tullis J (2001) Weakening and strain localization produced by syn-deformational reaction of plagioclase. Intl J Earth Sciences 90:136-148 Tullis J, Yund R (1999) Shear deformation of plagioclase aggregates: effects of composition and water. EOS Trans AGU 80:F1054 Tullis TE, Tullis J (1986) Experimental rock deformation techniques. In Hobbs BE, Heard HC (eds) Mineral and Rock Deformation: Laboratory Studies. The Paterson Volume. American Geophysical Union, Washington, DC, Geophys Monogr 36:297-324 von Kármán T (1911) Festigkeitsversuche unter allseitigem Druck. Z Verein Deutsch Ing 55: 1749-1757 Yamazaki D, Karato S (2002) Fabric development in (Mg,Fe)O during large strain, shear deformation: Implications for seismic anisotropy in Earth’s lower mantle. Phys Earth Planet Inter (in press) Zhang S, Karato S (1995) Lattice preferred orientation of olivine aggregates deformed in simple shear. Nature 375:774-777 Zhang S, Karato S, Fitz Gerald JD, Faul U, Zhou Y (2000) Simple shear deformation of olivine aggregates. Tectonophysics 316:133-152 Zimmerman ME, Zhang S, Kohlstedt DL, Karato S (1999) Melt distribution in mantle rocks deformed in shear. Geophys Res Lett 26:1505-1508

2

New Developments in Deformation Experiments at High Pressure William B. Durham University of California, Lawrence Livermore National Laboratory, P.O. Box 808 Livermore, California 94550

Donald J. Weidner Department of Geosciences State University of New York Stony Brook, New York 11794

Shun-ichiro Karato Departmentof Geology and Geophysics 319 Kline Geology Laboratory, Yale University New Haven, Connecticut 06520

Yanbin Wang Center for Adanced Radiation Studies University of Chicago Chicago, Illinois 60439 INTRODUCTION Although the importance of rheological properties in controlling the dynamics and evolution of the whole mantle of Earth is well-recognized, experimental studies of rheological properties and deformation-induced microstructures have mostly been limited to low-pressure conditions. This is mainly a result of technical limitations in conducting quantitative rheological experiments under high-pressure conditions. A combination of factors is changing this situation. Increased resolution of composition and configuration of Earth’s interior has created a greater demand for well-resolved laboratory measurement of the effects of pressure on the behavior of materials. Higher-strength materials have become readily available for containing high-pressure research devices, and new analytical capabilities—in particular very bright synchrotron X-ray sources—are now readily available to high-pressure researchers. One of the biggest issues in global geodynamics is the style of mantle convection and the nature of chemical differentiation associated with convectional mass transport. Although evidence for deep mantle circulation has recently been found through seismic tomography (e.g., van der Hilst et al. (1997)), complications in convection style have also been noted. They include (1) significant modifications of flow geometry across the mantle transition zone as seen from high resolution tomographic studies (Fukao et al. 1992; Masters et al. 2000; van der Hilst et al. 1991) and (2) complicated patterns of flow in the deep lower mantle (~1500-2500 km), perhaps caused by chemical heterogeneity (Kellogg et al. 1999; van der Hilst and Karason 1999). These studies indicate that while large-scale circulation involving the whole mantle no doubt occurs, significant deviations from simple flow geometry are also present. Two mineral properties have strong influence on convection: (1) density and (2) viscosity (rheology) contrasts. In the past, the effects of density contrast have been emphasized 1529-6466/00/0051-0002$05.00

22

Durham, Weidner, Karato & Wang

(Honda et al. 1993; Kellogg et al. 1999; Tackley et al. 1993), and the influence of rheology has been demonstrated by geodynamic modeling (Davies 1995; Karato et al. 2001). Rheological properties of the Earth’s mantle can be inferred from analyses of geodynamic data such as post-glacial crustal movement and gravity anomalies (Mitrovioca and Forte 1997; Peltier 1998). While such geodynamic inference provides an important data set on rheological structures of the Earth, there are serious limitations for these approaches. First, the resolution of post-glacial rebound data to infer mantle viscosity is limited to ~1200 km. Below that depth, there is little constraint for viscosity from post-glacial rebound (Mitrovioca and Peltier 1991). Furthermore, strain magnitude involved in post-glacial rebound is much smaller than strain magnitude involved in convection, and this raises the issue of effects of transient rheology (Karato 1998). Gravity data (e.g., the geoid) have better sensitivity to radial variation in viscosity in the deep mantle (Hager 1984), but they suffer from non-uniqueness (King 1995; Thoraval and Richards 1997). Therefore, experimental studies on mineral rheology remain a vital component in inferring mantle rheology. High-pressure rheology may also hold some of the keys for understanding deep earthquakes. Four classes of mechanism have been suggested over the years, including thermal runaway instabilities (Hobbs and Ord 1988; Ogawa 1987), instabilities accompanying recrystallization (Post 1977), dehydration embrittlement (Jiao et al. 2000; Raleigh and Paterson 1965), and instabilities associated with polymorphic phase transformations (see also reviews by Green and Houston (1995) and Kirby et al. (1996)). Again, the main reason uncertainty still exists is the poor knowledge of materials properties at high pressures. Experimental testing of the thermal runaway involved in plastic instability requires high rates of deformation, which are difficult to obtain in small samples; modeling would be a useful tool for this mechanism, but critical parameters for models (rheological properties of deep Earth materials) have been missing so far (Hobbs and Ord 1988; Karato et al. 2001). Well-controlled deformation experiments under deep mantle conditions are needed to provide greater understanding of the shearing instability associated with the phase transformations of olivine to wadsleyite and ringwoodite (Burnley et al. 1991; Green et al. 1990) and the possibility of faulting associated with dehydration of the dense hydrous magnesium silicates (Green 2001). Why is pressure important? Pressure is one of the most important variables in defining the ductile rheology of mantle minerals. Where the deformation mechanism is fixed, the effect of pressure on creep strength is given by

σ (P) /σ (0) ∝ exp(PV * /nRT) ,

(1)

where σ(P) is the creep strength at pressure P, V* is activation volume, n is the stress exponent, R is the gas constant, and T is temperature. For olivine, assuming an activation volume of V* = 15 cm3/mol, T = 1600 K and n = 3, σ(10 GPa)/σ(0) = 43. This is to be compared with the effect of water, which will change the creep strength of olivine by a factor of ~10 from dry to water-saturated (Hirth and Kohlstedt 1996; Karato and Jung 2002 in press) or partial melting that will affect creep strength by no more than a factor of 2 (Kohlstedt and Chopra 1994). However, the pressure effect is small at the low pressures where studies are made, e.g. σ(0.5 GPa)/σ(0) = 1.2. Therefore, high-pressure studies are necessary in order to determine the flow properties in the deep Earth. Such large effects of pressure on the flow laws can result in changes of dominant flow mechanisms with depth in the Earth. Karato and Wu (1993) suggest that dislocation flow is overtaken by diffusion flow as the creep mechanism for olivine within the upper

Developments in Deformation Experiments at High Pressure

23

mantle. Their reasoning is based on the differences in the pressure dependence, i.e., in the value of V* in Equation (1), of the two flow laws. Pressure is an agent for phase transformation within the Earth. These high-pressure phases, with different crystal structures, will also differ in their mechanical properties. Rheology of high-pressure phases needs to be investigated within the stability field of the mineral. Because of equipment limitations, the flow properties of minerals such as ringwoodite or majorite, which set the strength characteristics of the transition zone, remain virtually unknown. The effects of water can be well characterized only by high-pressure experiments. At low pressures (P1 GPa), the creep strength of olivine increases with pressure even at water-saturated conditions (Karato and Jung 2002 in press). Therefore results that can be extrapolated to infer creep strength (viscosity) at deep portions (>30 km) of Earth can be obtained only by quantitative experiments at pressures higher than ~1 GPa. Terminology related to strength and deformation. Usage sometimes varies among research disciplines, so let us define a few key terms, taking the language of rock mechanics as a basis (e.g., Jaeger and Cook (1976)). It is assumed that the meaning of the stress (σij) and strain (εij) tensors are understood. Deformation (used already in the title without definition) refers loosely to any process that results in strain. We take positive values of stress and strain to indicate compression, following geologic convention. Pressure, hydrostatic pressure, and mean normal stress are all equivalently the trace P of the stress tensor, namely,

σij = σ ′ij + P ,

(2)

where σ ′ij is the deviatoric stress tensor, i, j = 1, 2, 3, representing orthogonal directions. Confining pressure Pc is the hydrostatic stress generated by laboratory devices to simulate geologic overburden. Although P and Pc are different whenσ ′ij is non-zero, the difference in most geologic settings is small and the distinction is very often ignored. Different terms for the same physical quantity are sometimes used to connote different situations. Strain can be divided into two categories: elastic, the recoverable portion, which owes its existence to the state of non-zero σij, and inelastic, the permanent strain that remains forever after stress has been removed. Classic rock mechanics usage recognizes two categories of deformation that result in inelastic strain: brittle and ductile. Brittle deformation is associated with discontinuities of displacement and sudden loss of strength (a term to be defined shortly) and is generally not volume conservative. Ductile deformation is associated with finite displacement gradients, retention of strength, and volume conservation at scales above the nm. Note that there are important strainproducing phenomena that do not fit well into either category, in particular those that involve polymorphism (phase change) or chemical reaction. They may be ignored for the purposes of this review. The strength of a material (in specific circumstances called the yield strength, ultimate strength, etc.) is the maximum stress the material can support under the stated environmental conditions. Flow and failure are closely related terms for the state that exists the moment that applied stress matches strength for ductile and brittle materials, respectively. The term plasticity requires special note. It was originally defined for metals as essentially equivalent to the phenomenon of flow, just defined, but was defined in terms of the yield envelope (Hill 1950) without qualification as to ductile or brittle yield. While metallurgists and most rock rheologists have retained this meaning for the term plasticity

24

Durham, Weidner, Karato & Wang

(e.g., Evans and Kohlstedt [1995] and Poirier [1985], many in the computational branches of geology have taken plasticity to mean any yield process, whether volume conservative or not. In this review, we follow tradition and use the terms ductile and plastic interchangeably. Here we focus exclusively on ductile deformation, so we use the terms rheology (flow of materials under stress), deformation, and flow interchangeably. The rheology of a material is described by a constitutive law of the type εÝij = Af (σ ij′ ) , A = A(P, T, …) (3) where all relevant state variables are included in the term A. Engineering the containment of very high pressures in the laboratory generally requires high symmetry in the sample assembly. As will be seen below, two basic types of high-symmetry strain environments exist: axisymmetric and rotational. A shorthand form of Equation (3) has evolved for each of these geometries and we will use that shorthand here. For axisymmetric deformation (whose two-dimensional analog is called pure shear), the values of the three principal stresses are σ1 ≠ σ2 = σ3, and we simplify the strain rate tensor to a scalar εÝ, which we call the strain rate or shortening rate. The deviatoric stress tensor can be characterized by the scalar σ = σ1 – σ3, which we call the differential stress or simply stress. For rotational deformation, the strain is approximately simple shear, and Equation (3) becomes a scalar relationship between the shear strain rate, γÝ, and the shear stress, τ. Despite the apparent simplicity of this language of strength and deformation, one should always keep two things in mind: (1) the terminology is not universal, and (2) the simple language belies the flow of rocks in nature, which can be counted on to be both inhomogeneous and low symmetry. A BRIEF HISTORY OF HIGH-PRESSURE APPARATUS To 5 GPa: Cylindrical devices Confining pressure in rock mechanics testing serves several purposes: (1) simulating geologic overburden, (2) activating mechanisms of flow and fracture that depend on both normal and shear components of stress, and (3) suppressing brittle deformation where one is interested in gathering information on purely ductile phenomena. The scientific needs for doing measurements under pressure are illustrated in detail in several other contributions to this volume. These needs have driven the development of experimental pressure systems since the early days of rock mechanics. The earliest experimental machines for deforming rocks under elevated confining pressure were some version of a piston pushing on a cylindrical rock held fast in a cylindrical container. A solid or fluid medium within the container, or even walls of the container itself, provided confining pressure to the rock sample. The reader may find an early history of such machines in Tullis and Tullis (1986), for example. The direction of apparatus development after 1900 was influenced primarily by P.W. Bridgman, who made such rapid advances in the design of hydrostatic equipment that the only reasonable path for rock mechanics to follow was to adapt deformation to Bridgman’s vessels. Bridgman, who conceived the dynamic piston seal still commonly used, was involved in that adaptation (Bridgman 1952c), although he was no longer alone in his efforts. Continuing to this day, virtually every new deformation rig is some adaptation of a hydrostatic design. The backbone of the history of deformation machines is mostly a history of pressure vessels. The technological breakthrough behind modern gas and liquid pressure vessels was

Developments in Deformation Experiments at High Pressure

25

the Bridgman unsupported area packing (Bridgman 1914), a geometry that intensified the vessel’s own pressure at the point of seal and made it possible to “reach without leak any pressure allowed by the mechanical strength of the walls of the containing vessels” (Bridgman 1952b). Bridgman regularly worked at pressures in excess of 1 GPa and occasionally reached 2 GPa with his fluid vessels (Bridgman 1935b). The application of the Bridgman-seal vessel to rock deformation led to the creation of what is commonly called in rock mechanics the “triaxial” apparatus (which is somewhat inappropriate, because the stress environment in the fluid pressure medium can be no lower than axisymmetric). Griggs, who learned the techniques of high pressure from Bridgman and shared his “gift for gadgets” (Rubey 1972), initiated this application by developing a version of the unsupported area packing around a solid moving piston (Fig. 1), thereby allowing sample shortening to be equated (with appropriate corrections for elasticity) to piston displacement (Griggs 1936). Griggs (1936) added a second innovation to the design, a pressure-compensating double piston that automatically kept pressure constant even as the sample was shortened (Fig. 1). Further development of the triaxial rig was aided by advances such as the controlled-clearance (Newhall) packing to reduce piston friction (Handin 1953; Turner et al. 1956); internal heating to achieve higher sample temperatures (Griggs et al. 1960); and the internal force gauge, which removed piston friction from the measurement of differential force on the sample (Heard and Carter 1968; Paterson 1970). This list is by no means exhaustive. A notable recent development is the proliferation in many of the world’s most renown rock mechanics laboratories of one particular triaxial apparatus, the Paterson rig, which is a 0.3-GPa ready-touse system complete with actuation and internal force gauging for both axial and high-displacement torsional deformation (Paterson 1990; Paterson and Olgaard 2000). As of this writing, there are ten such vessels in use in North America and Europe. Practical matters related to sealing a high-pressure fluid have generally limited triaxial work to about P = 1 GPa, with most triaxial machines, in fact, having maximum design pressures of 0.3-0.5 GPa (Paterson 1978). At higher pressures, the solid pressure medium piston-cylinder apparatus is used. The solid-medium press used most widely for geological applications is the Griggs apparatus (Griggs 1967), adapted from the hydrostatic solid-medium pistoncylinder apparatus (Boyd and England 1960), and subsequently improved to provide a stress resolution of a few MPa (Borch and Green 1989; Green and Borch 1989; Tingle et al. 1993). Since a solid medium has a finite shear strength, measurements are limited to materials

Figure 1. The Griggs triaxial deformation apparatus, including the yoked pistons to hold pressure constant during deformation. As piston P1 moves against the sample, piston P2 automatically withdraws from the vessel at the same rate, holding the pressurized volume constant. After Griggs (1936), Fig. 2; reproduced with permission.

26

Durham, Weidner, Karato & Wang

that are stronger than the medium, so very low strain rates and temperatures near the sample melting point must be avoided, but pressures of several GPa are now easily accessible. The Griggs apparatus has been regularly used to 1.5-1.8 GPa, and to a maximum pressure of 4 GPa. A recent development is an apparatus built by Getting (Getting 1998; Getting and Spetzler 1993) that combines aspects of the solid-medium and fluid-triaxial deformation machines. Getting’s is a piston-cylinder device that features precision packings that are sufficiently tight to contain solid argon (an exceedingly weak solid) but sufficiently reproducible in their frictional behavior that much of the uncertainty in stress introduced by piston friction can be accounted for by calibration. Getting (1998) claims pressure uncertainty at P = 3 GPa at the extraordinary level of a few MPa. If this resolution applies also to the sum of σ1 + P, as expected, it may mean that rock strengths can be measured with a resolution that is comparable to that of many triaxial gas rigs that operate at far lower pressures. The practical pressure limit for all cylindrical machines is around 5 GPa. While the pressure capacity of a hollow cylinder (or any shape, for that matter) is a function of wall thickness and therefore virtually unlimited (Eremets 1996), the same is not true for the piston. As one approaches P = 5 GPa, the stress required to drive the piston into the pressure cylinder (which must be 5 GPa plus the stress to overcome packing friction plus the strength of the sample) causes pistons constructed of even the strongest of steels to creep. Pistons will swell and seize in a short amount of time. Note that this 5-GPa limit exists for hydrostatic piston-cylinder vessels as well. Bridgman reached higher pressures in hydrostatic vessels by confining the entire piston and cylinder in a larger vessel, thus raising σ3 on the piston to the pressure in the outer vessel and allowing the piston to apply correspondingly more stress to the inner vessel without exceeding its strength limit. His vessel-inside-vessel multi-staged devices have been used to reach pressures of at least 10 GPa (e.g., Bridgman (1942; 1948)). Because of the complexity of those vessels, however, they are not well suited to being instrumented for deformation studies. 5 GPa and above: Anvil devices A much simpler method for achieving an elevated σ3 is to give the pistons a tapered shape. (The change from rectangular to trapezoidal section also merits a name change: from piston to anvil.) Now a component of the force applied in the σ1 direction at the larger end of the piston/anvil is directed in the σ3 direction at the smaller end (platen face), making σ3 > 0 at the platen face. The drawback of this solution is, of course, that one cannot expect to contain the pressure medium over large displacements in a conventional cylindrical pressure vessel if the sides of the (moving) anvil are tapered. Sealing against a moving tapered surface can be achieved for small displacements through use of a crushable gasket material; so if one is satisfied with anvil displacements that are less than gasket thickness, one can reach high values of σ1 at the platen faces without exceeding the failure criterion of the anvils. The first application of this principle is generally credited to Bridgman’s (Bridgman 1952a) opposed-anvil device, which is capable of 10 GPa, although Bridgman used the same principle many years earlier in confining samples in a 5-GPa shear apparatus (Bridgman 1935a). It is also the underlying principle of the Drickamer cell (see below). Note that the original purpose of anvil displacement in these designs was pressure generation rather than sample deformation, although as we will see below, the possibility of the latter was not ignored. A slight variation, which allows for larger volumes at a given pressure, is to combine the function of pressure container and pressure generator by clustering several anvils in a quasi-spherically symmetric pattern. These are called multianvil devices. Four-anvil

Developments in Deformation Experiments at High Pressure

27

tetrahedral (Hall 1958; Houck and Hutton 1963; Lloyd et al. 1959), six-anvil cubic (Carter et al. 1964; Houck and Hutton 1963; Osugi et al. 1964), and eight-anvil octahedral (Kawai and Endo 1970; Walker et al. 1990) have been constructed. The eightanvil devices are multistage, with the eight anvils of the final stage contained within the working volume of an intermediate cubic stage. They are often referred to, therefore, as 6/8 devices. Hydraulic actuation of the several anvils was sometimes independent or in opposed pairs (Carter et al. 1964; Hall 1958), but better symmetry of anvil displacement was achieved, along with higher pressures, when the anvils were driven by wedge-shaped guide blocks that then required only one hydraulic actuator (Lloyd et al. 1959). There are at least two versions of the wedge-type cubic multianvil device: a split-sphere device (Houck and Hutton 1963) and the DIA (Osugi et al. 1964; Shimomura et al. 1985); the latter will be describe below in detail. Wedged guide block machines achieve high pressures more reliably than machines with multiple actuators, and for that reason they have proliferated, especially the DIA and 6/8 devices (e.g., see review by Onodera (1987)). Note, however, that their design function is exclusively hydrostatic pressure generation. Advances in high-pressure rheology required that the quasi-isotropy within the pressurized volume be broken; those advances are the main subjects of this review. Another reason for the recent proliferation of multianvil devices is a serendipitous advantage they have over pressure cylinders: optical and/or X-ray transparency along a line of sight leading directly to and from the sample (Fig. 2). In the case of diamond (or sapphire) anvil cells, the anvils themselves are transparent. For devices where anvils are opaque, lines of sight are always available in the narrow gaps between anvils, and the gaskets compressed in those gaps can be made of X-ray-transparent weak material (e.g., pyrophyllite). The analytical potential of high-energy X-rays and their recent availability at very bright synchrotron sources has been the motivation for development of a new generation of deformation machines that have the pressure capacity of anvil devices and the lowered symmetry of deformation machines. MEASUREMENT METHODS AT HIGH PRESSURE Stress measurement Measurement of deviatoric stress is key to quantifying the rheological properties of a material. Stress provides information about strength relative to the laboratory time scale. Stress, measured in conjunction with ductile strain, can yield flow laws within the context of the state of the sample. The crucial ingredient to facilitate such studies is the development of a stress meter, or piezometer. In experiments operating below 5 GPa, the deforming force is applied directly to the sample and various strategies are designed to measure the magnitude of this deforming force. Stress is then given as the ratio of force to sample area. Typically, a load cell placed in series with the deforming piston defines this force along the deforming column. Combinations of friction between the force gauge and the sample and forces that support the piston, such as in the multianvil system, continually degrade the ability to relate a measured force to sample stress. These problems begin to interfere with the resulting accuracy at 0.3 GPa and completely overwhelm the measurements by 5 GPa. A revolutionary technique for stress and strain measurements under high-pressure (and temperature) conditions in situ has now become feasible using high-energy X-rays generated by a synchrotron radiation facility. These new tools are just now being explored and their limitations defined. The exceptional quality of these tools rests in the fact that they are directly monitoring the sample. Stresses are measured in the sample.

28

Durham, Weidner, Karato & Wang

Figure 2. Sample assembly in the DIA cubic apparatus, shown with three of the six anvils removed and with the sample assembly itself, i.e., the cube, shown in cross section. Analytical synchrotron X-rays enter and exit along a horizontal line through the sample and normal to the section as shown here. The X-rays pass through the vertical gaps between anvils. Drawing by M. Vaughan.

Different positions in the sample can be isolated to test for stress uniformity. Measurements of stress and strain can be time resolved with a precision of about one minute. Strain is also obtained by images of the sample. Distribution of strain with position and time can be defined. All of this is done with a current accuracy of a few tens of MPa in differential stress and 10-4 in strain. Strategies for measuring stress are found through use of the X-ray diffraction or pressure-sensitive probes (such as ruby fluorescence), and are being made available by exploiting powerful synchrotron X-ray sources. The basic principle of X-ray piezometers is that lattice strain is sensitive to elastic strains and not to plastic strains. Changes in distances between atoms are expected to balance stress fields that are present. These stress fields may be created by dislocations or external forces, but departures from the lattice spacing that the material experiences at zero pressure and stress are sensibly related to the stress field within the sample. Furthermore, the coupling between the strain and stress is expressed by the elastic moduli. There are two styles of X-ray piezometers, one based on stress gradient and the other based on strain anisotropy.

Developments in Deformation Experiments at High Pressure

29

Stress gradient. Where the inertia term can be neglected, variation in the stress field is restricted to obey the mechanical equilibrium relations: 3

∂σ

∑ ∂x ijj = 0 j= 1

Three equations, one for each value of i, represent the vector force balance equations. The first of these equations is:

∂σ 11 ∂σ 12 ∂σ 13 + + =0 ∂x1 ∂x 2 ∂x 3 Thus, if there is a variation of the normal stress, σ11, with x1, there must also be a variation of the shear stresses, σ12, in the x2 direction and/or a variation of σ13 in the x3 direction. The peak magnitude of the shear stress is related to the variation of the normal stress (or pressure) and the length scales of two types of stress. This has been exploited in the diamond anvil cell geometry by measuring the radial gradient of pressure to deduce the axial (along the axis of the diamonds) variation of shear stress. This is illustrated in Figure 3. The shear stress, τ, vanishes along the center plane of the sample and along the central axis of the cell where ∂P/∂r is zero. The maximum shear stress is at the samplediamond interface and is given by τmax=(∂P/∂r)max d/2 where d is the thickness of the sample. This technique was developed by Sung et al. (1977) and was further used by Meade and Jeanloz (1988a,b; 1990). These experiments relied on ruby dust distributed across the sample as a pressure marker for delineating the

Figure 3. Illustration of stress field in a diamond-anvil cell. The radial distribution of pressure, illustrated at the top, is coupled to the shear stress distribution along the axial direction. Force couples indicate the sense of the shear stress distribution within the cell.

30

Durham, Weidner, Karato & Wang

pressure gradient. As Wu and Bassett (1993) illustrate, the pressure marker can also be an X-ray diffraction standard such as gold. The objective is to measure the radial pressure profile to define the shear-stress distribution. Often, the peak shear stress is taken as the yield point of the solid. Stress gradients coupled with deviatoric stress affect the diffraction peak shape as described in detail by Weidner(1998) and Weidner et al. (1998). In this case, heterogeneities in the stress field broaden the diffraction peaks. Since the length scale of stress variation is the same for normal and shear stress, then the magnitude of the heterogeneity for all stresses is similar. Weidner et al. (1998) model the relationship between the observed strain (that creates the broadened diffraction peak) and the stress magnitude and conclude that Young’s modulus is the appropriate coupling elastic modulus, or σdifferential = E εdifferential , where the differential strain is deduced as the “strain broadening” contribution to the peak profile. Since for the peak-broadening measurement, the full-width, half-height of the diffraction peak is used to define the strain broadening and represents roughly twothirds of the scattering diffraction planes, then most of the sample must be experiencing the inferred shear stress. Deviatoric stress can originate from dislocations, from heterogeneities in material or properties, or any number of sources. Peak-broadening analysis is often used in commercial applications to determine stresses induced by processing. For example, internal stress generated by welding is often analyzed by this method. In high-pressure studies, the stress is most commonly generated by compressing powdered samples (e.g., Chen et al. 1998). As a rheological tool, this approach resembles an indentation experiment where the indenter is the same material as the sample. Weidner et al. (1994) studied the high-temperature and -pressure strength of diamond with this method. Chen et al. (1998) studied plastic deformation of olivine, wadsleyite, and ringwoodite. Flow can also be studied with this piezometer, with the constraint that this becomes a relaxation experiment (Weidner et al. 2001). Strain anisotropy. A uniaxial stress will introduce elastic strains in the sample. The strain parallel to the axis of compression will generally be larger than the strain perpendicular. The material’s elastic moduli quantitatively define the relationship between these strains and the imposed stresses. X-rays sample the distance between lattice planes (called dspacing) whose normals are parallel to the diffraction vector, which is the bisector between the incident X-ray and the detector. The lattice spacing is insensitive to strain history (plastic strain) but reflects the elastic strain field. Figure 4 illustrates a Debye ring that would be observed from a powder sample in a stress field using a monochromatic Xray beam and an area detector. The lattice spacing is related to the distance from the center of the image to the Debye ring through Bragg’s law. For the angle Ψ of zero, the lattice spacing reflects that measured parallel to the x1 direction. For Ψ of ninety degrees, the X-rays are sampling grains aligned parallel to the x2 direction. The strains deduced from these measurements compared with ambient conditions are related to the stresses by 3

3

ε ij = ∑ ∑ sijklσ kl , k =1 l=1

where s is the elastic compliance tensor. In the actual case, more than one set of diffraction planes will produce Debye rings similar to the one illustrated here. Each will be approximately circular in shape, but distorted by the stress field. The amount of distortion may be different if the elastic

Developments in Deformation Experiments at High Pressure

31

Figure 4. X-ray diffraction line on a two-dimensional detector from a sample under stress. The distance from the center to the ring defines the scattering angle, 2θ , and hence the lattice spacing through Bragg’s law. The presence of the stress field distorts the Debye ring from its normally circular shape. The usual stress field is axisymmetric, with σ2 = σ3. The diffraction vector at ψ = 0 lies at an angle θ from the direction of σ1. θ is typically only 3 to 4 degrees.

compliance tensor is not isotropic. To evaluate this in detail, we need to be careful to account for the crystal orientation when defining the compliance tensor. Thus, in a nonhydrostatic stress field, crystallographic planes with different orientations change their spacing by different amounts. For cubic crystals, the relationship between the Ψ dependence of strain and the differential stress, Δσ, is given by: εhkl(Ψ) = Δσ /3 {1 – 3cos2 Ψ} {(S11 – S12) [1 – Γ(hkl)] + S44 Γ(hkl) / 2} Each diffraction peak corresponds to a set of lattice planes defined by the Miller index, hkl. In an anisotropic crystal, the elastic moduli will depend on the particular set of planes that are considered. The function, Γ(hkl), varies from 0 to 1 and accounts for this variation as it multiplies the different elastic compliance moduli, Sij. Thus, for a particular diffraction plane, hkl, the right hand bracket contains a constant value. Then ε varies in a fixed manner with the angle, Ψ. The above expression is based on the assumption that each grain experiences the same stress field, or the Reuss condition. A similar relation can be derived by assuming that each grain satisfies the Voigt condition of iso-strain. In this case the equation is altered by replacing the elastic shear compliances, S44 and (S11 – S12) by their Voigt equivalent. This relation for cubic symmetry is generalized by Singh et al. (1998) for all crystal symmetries. This phenomenon gives rise to a couple of strategies for a piezometer. Stress has a manifestation even for a single value of Ψ. For an elastically anisotropic material, the strain for a given crystallographic plane will depend on the orientation of crystal with respect to the applied stress and the stress magnitude. Therefore by measuring the strain for several diffraction peaks corresponding to different orientations, one can determine the stress magnitude using the known elastic constants of the crystal. Such a measurement can now be done in a multianvil high-pressure apparatus using a synchrotron X-ray source (Weidner et al. 1992). The accuracy relies on knowing the elastic anisotropy at the pressure/temperature conditions of the experiment. A more robust strategy comes from mapping out the d spacing as a function of Ψ. The values of the elastic moduli are still required, but the details of anisotropy are less critical to yielding precise measurements. With a precision of 10-4 for the lattice spacing

32

Durham, Weidner, Karato & Wang

and a typical elastic modulus of 200 GPa, it should be possible to resolve differential stresses of ~20 MPa. Singh has reduced the elasticity equations to a closed-form solution for several crystal symmetries and has given a concise summary (Singh et al. 1998). This methodology was first applied in a diamond-anvil cell by Kinsland and Bassett (1977) by passing the X-ray beam through a Be gasket, perpendicular to the axis of the diamonds. The multianvil system has not been accessible by this technique because the anvils themselves cast a shadow and limit the range of Ψ where the diffracted signal can be observed. However, we have recently experimented with cubic BN anvils. These are nearly as hard as diamond but are about an order of magnitude less expensive. Furthermore, they are transparent to X-rays. Thus, by using these anvils it is now possible to observe diffraction spectra in any plane relative to the incident beam. Here we have illustrated this tool with monochromatic X-rays and angle dispersive measurements. Energy dispersive methods are equally valid for these measurements and are sometimes advantageous. Spectra need to be collected for different values of Ψ through collimators that fix the two-theta value. Weidner’s group has designed and used a conical slit system for this purpose as illustrated in Figure 5. Two concentric cones, whose angle is the desired diffraction angle for white, energy dispersive diffraction, create the slit itself. Solid-state detectors with energy discrimination are placed at specific values of Ψ behind the slit system. Each detector is calibrated independently. Lattice spacings are used as illustrated above to define the differential stress.

Figure 5. Conical slit assembly in use at National Synchrotron Light Source. The cone defines the two-theta angle and is constructed by two concentric cones with this angle. The X-rays that pass between the two cones originate from the sample at that angle as all other X-rays are blocked by the cones. Multiple energy discriminating detectors, located behind the slit assembly, record the X-ray signals. A YAG crystal located inside the slit assembly fluoresces in the X-ray beam, emitting visible light where an X-ray photon passes. The visible light is magnified by a microscope assembly and recorded by a digital camera.

Strain rate measurement Synchrotron X-rays can also be used to measure sample length. Direct images of the sample can be obtained using an incident beam whose dimensions are larger than the sample. Platinum or gold foils above and below the sample are easily viewed. The X-ray image is obtained by projecting the X-ray onto a fluorescent screen that is viewed with a CCD camera though a magnifying system. A typical image is shown in Figure 6. The

Developments in Deformation Experiments at High Pressure

33

horizontal lines in this image are gold discs placed at the ends of the sample. The black edges are the shadows of the opaque tungsten carbide anvils. Cubic boron nitride anvils are transparent and can be used to view the entire sample because they do not cast such a shadow. Figure 5 illustrates the position of the microscope for these measurements. The fluorescent screen is located on-axis inside the conical slit system. We determine that, by comparing two images, strains of 10-4 can be measured. For images taken 100 seconds apart, this allows resolution of strain rates of 10-6 s-1.

Figure 6. X-ray shadowgraph of MgO sample during compression. The horizontal dark lines are images of gold foil that bound the sample whose length is ~1 mm. The horizontal dimension is defined by the opening between the high-pressure anvils. The shadow across the top is caused by a platinum foil wrapped around the end-plug.

The X-ray image yields the total strain of the sample, which is the sum of the elastic and plastic strain or: εob = εel + εpl In an experiment both the elastic and plastic components may vary with time. Thus, to deduce the plastic strain rate, it is necessary to account for the variations of all strain components. The elastic strain, however, is reflected in the diffraction pattern and can be represented by a unit cell length, a. This cell-edge length is obtained from analysis of all of the diffraction peaks recorded with a diffraction vector parallel to the sample axis. In elastically isotropic samples, there is no ambiguity in defining this value. Anisotropic samples in a differential stress field will exhibit lattice spacings that reflect the aggregate properties. In this case, an average cell dimension based on several diffraction peaks should be obtained. We define the Kung ratio, Rk, as: Rk = (d/d0) / (a/a0) where d is the length of the sample measured with an image and a is a unit cell length measured from diffraction; the zero subscript refers to a reference value such as ambient conditions. This ratio, which is normalized to one, represents the length of the sample in the units of the cell dimension. This number is proportional to the number of unit cells that make up the length. The Kung ratio remains constant if all deformations are elastic. Thermal expansion, pressure, and stress will not change the Kung ratio if the sample responds elastically. Plastic strain changes the Kung ratio. In this case the number of unit cells that define the length changes. Thus, the Kung ratio can be used as the proxy for plastic strain, or: εpl = [Rk(2) – Rk(1)] / Rk(1) where (1) and (2) represent two states of the material. Strain is not always uniform in the sample. Visual tools as described here provide opportunity to map the strain field as a function of space and time during the experiment. Vaughan et al. (2000) demonstrate that one can embed strain markers in a sample during preparation and view the position of these markers during the experiment. Figure 7, from this work, shows a grid made from a TEM gold sample grid, in the sample at different P, T conditions. The sample was a sintered olivine cylinder. The cylinder was sliced into

34

Durham, Weidner, Karato & Wang

two halves, the grid inserted, and the sample halves placed back together. The sample assembly included hard end-plugs in a DIA apparatus, so as to produce shortening during compression. The grid size is initially about 70 microns. With these images, it is possible to map the strain as a function of position in the sample and as a function of time by comparing successive images.

Figure 7. A TEM gold grid embedded on the central plane of a cylindrical sample. Images at different conditions illustrate different amounts of strain. The initial pressurization distorted the grid because surfaces of sample assembly parts were not perfectly flat. Subsequent strain can be mapped by following specific strain markers.

MODERN TECHNIQUES FOR DEFORMATION AT HIGH PRESSURES In current technology, deformation at P > 10 GPa is the realm of anvil devices. We describe in detail four methods currently in use or in active development: (1) sample assembly modification to hydrostatic multianvil devices, (2) sample assembly modification to the DAC, (3) the D-DIA, and (4) the RDA. Sample assemblies Sample assemblies used in these machines share several common features of those used in hydrostatic devices, serving common requirements such as high pressure, high temperature, and control of chemical activity in the sample. With the exception of the specialized environment of the DAC, these common features include a pressure medium, such as soft fired pyrophyllite, ZrO2, MgO/5%Cr2O3, castable ceramic or epoxy; a furnace, for example, metal foil, LaCrO3, or graphite; and a soft medium surrounding the sample itself, for example, BN, MgO, NaCl, or glass (the latter two occasionally as liquids). Descriptions of such assemblies can be found in, e.g., Kawai and Endo (1970), Liebermann and Wang (1992), Walker(1991), and Weidner et al. (1992). For control of chemical environment the sample can be encapsulated along with buffering compounds (see e.g., Rubie et al. 1993). Non-hydrostaticity is introduced by introducing hard platens that impart shear or

Developments in Deformation Experiments at High Pressure

35

normal stresses to the sample. Several of the methods for achieving this are compared in Figure 8. They provide axisymmetric deformation in the multianvil devices (Figs. 8a-8c), simple shear in the multianvil (Fig. 8d), axisymmetric deformation in the DAC (Fig. 8e), and rotational (approximately simple shear) deformation in the RDA (Fig. 8f). We compare these assemblies briefly here and in more detail below.

Figure 8. Starting (P = 0) configurations of sample columns used for deformation experiments at very high pressures. For simplicity, details such as sample encapsulation, thermocouples, and furnace electrodes are not shown. (a) Stress relaxation. The sample is loaded elastically during pressurization. The furnace and outer and inner pressure media shown here also apply to (b)-(d). (b) Constant stress/strain rate tests. The porosity within the crushable ends of the column takes up most of the column shortening during pressurization. (c) Specialized version of (b), with a fluid in the column assuring strictly hydrostatic loading of the sample during pressurization. (d) Simple shear, high strain. (e) Column for deformation in the diamond-anvil cell. Use of a separate piston is optional, i.e., the diamond anvils can also act as pistons. In some applications, a ruby piston can also serve as a differential stress gage. (f) Assembly for the rotational Drickamer apparatus, also shown in Figure 13.

The adaptations to the DIA and 6/8 discussed in the next section use a column such as shown in Figure 8a, in which the source of deviatoric stress in the sample is the closure of the anvils during initial pressurization. At the scale of Figure 8a, the shortening of the column (elastic plus inelastic) is approximately 20% of the column length (Durham and Rubie 1998), so the “crushable” material at the ends of the column, present to protect thermocouple wires and furnace electrodes, has a porosity somewhat less than 20% of the column length. To avoid deforming the sample during pressurization in devices capable of deforming at fixed pressure (currently only the D-DIA) one uses a column such as in Figure 8b, where the length of crushable material is sufficient to absorb all anvil displacement during pressurization When a strictly hydrostatic environment is required during pressurization, such as when the sample is a fragile single crystal, it is also possible to create a liquid cell around the sample before the pressurizing load is applied (Fig. 8c) (Durham, unpublished results).

36

Durham, Weidner, Karato & Wang

Figure 8d shows the method for converting axial displacement to high-strain, simpleshear deformation in the multianvil cell (Karato and Rubie 1997). As discussed below, this design was first introduced as a method for imposing larger strains in the stressrelaxation mode of the hydrostatic multianvil device. However, it can also be used in the D-DIA, presumably with slightly more crushable material at the ends, to carry out constant stress or constant strain-rate tests to very high simple-shear strains. The very simple deformation assembly for the DAC (Fig. 8e) reflects the greatly reduced available volume and the fact that laser heating can be used in place of a resistance heater to achieve local heating of the sample. Finally, the assembly for the RDA (Fig. 8f) is the rotational analog to the axisymmetric columns in Figures 8a-8c. Because rotational deformation and pressurization are entirely decoupled in the RDA, there is no need for crushable material in the cell. Modifications to the sample assembly in a multianvil press Fujimura et al. (1981) developed one of the first techniques for converting the environment of the 6/8 device from hydrostatic to non-hydrostatic, thus making the 6/8 device a sort of deformation apparatus. By making the elastic character and/or mechanical strength of a sample assembly anisotropic, one can create a deviatoric stress in the assembly (e.g., Fig. 8a). This deviatoric stress can be relaxed by plastic flow upon heating. This technique has been used in numerous studies (Bussod et al. 1993; Durham and Rubie 1998; Fujimura 1989; Green et al. 1990; Karato et al. 1998; Karato and Rubie 1997; Weidner et al. 2001; Weidner et al. 1998). One typically achieves the anisotropic strain by embedding a rigid sample column in the compliant pressure medium. As the assembly is pressurized at room temperature (where most relevant geologic materials are very strong), a large deviatoric stress is generated on the sample. When temperature is increased, the sample softens and begins to flow, and the stress gradually relaxes. The analysis by Karato and Rubie (1997) (see also Durham and Rubie [1998]) showed that the mode of deformation in most cases is this “stress relaxation,” i.e., the magnitude of deviatoric stress changes significantly within a single experiment. This technique has several limitations. First, pressurization and plastic deformation are not completely separated; therefore, deformation during pressurization likely occurs unless special care is taken in sample assembly to absorb the initial-stage shape change of sample assembly (Durham and Rubie 1998; Karato and Rubie 1997). Second, deformation by this method is “stress relaxation,” and the interpretation is complicated because of the change in stress that could cause the change in deformation mechanisms. Third, the amount of inelastic displacement is limited to the amount that can be imposed elastically during pressurization, usually well under 100 μm. By using a thin sample sandwiched between two pistons cut at 45°, Karato and Rubie (1997) were able to convert this very small displacement to relatively large strains (γ ≈ 1-2) (Fig. 8d), yet the maximum strain is not large enough to obtain steady-state fabrics. Quasi-constant displacement rate tests can also be made with multianvil apparatus through the continuous movement of the hydraulic ram (Bussod et al. 1993; Green et al. 1990), although one must realize that pressure is increasing steadily and significantly in such tests. The multianvil deformation technique has been applied to ~25 GPa, ~2000 K with a sample dimension of ~1-2 mm diameter and ~2-4 mm long, or a thin (~0.2 mm) disk sample with a similar diameter (Chen et al. 1998; Cordier and Rubie 2001). Sacrificing pressure for sample size, Green et al. (1990) deformed 3-mm diameter x 6-mm length samples in a 6/8 device. The major advantage of this technique as compared with a diamond-anvil cell technique is that rheological properties can be investigated under more homogeneous high temperature and pressure conditions with a better-controlled

Developments in Deformation Experiments at High Pressure

37

chemical environment. Also, because relatively large samples can be used, microstructural evolution during deformation and its effects on rheology can be investigated (Green et al. 1990, 1992; Karato 1998a; Karato et al. 1998). The limited plastic deformation is a serious shortcoming of the method. The displacement available is generally not sufficient to probe important rheological questions about lattice preferred orientation and strength in the steady state. Second, because the stress magnitude changes during an experiment, dominant mechanisms of deformation may change in a single run, making interpretation of mechanical data and microstructures difficult. This last point, however, may turn out to be an advantage for the study of rheology at a very small strain-rate (small stress) that is relevant to Earth. Diamond-anvil cell The highest pressures obtainable in the laboratory are in the diamond-anvil cell. This device loads the sample uniaxially and, thus, naturally produces a deviatoric stress environment. The first examinations of differential stresses in the diamond cell were motivated by the need to produce a hydrostatic environment. Use of the diamond cell as a deformation device was pioneered by Kinsland and Bassett (1977) and Sung et al. (1977). The former used X-ray transparent gaskets and passed the X-ray beam through the gaskets, perpendicular to the diamond-cell axis. Stress was measured from the shape of the Debye rings as discussed above. The measurements provided the first estimates of the room temperature strength of MgO at high pressures. Sung et al. (1977) developed the methodology for using pressure gradients to deduce shear stress. In this case, a sample is sandwiched between two single crystals of diamond, and radial distribution of pressure is determined by measuring the shift of fluorescence lines of ruby crystals located at various points in the sample space. Then, assuming that the pressure gradient is supported by the sample strength, one can estimate the sample’s strength from the equation for force balance. This technique has been used extensively by Meade and Jeanloz (1988a,b; 1990) at room temperature to the pressure of ~40 GPa. Chai and Brown (1996) developed a diamond-cell piezometer based on the splitting of the ruby fluorescence lines. They find that the R2 line reflects the average stress, that the splitting between the R1 and R2 lines is sensitive to the differential stress, and that the character of the sensitivity depends on whether the load is applied parallel to the c-axis or the a-axis. This phenomenon suggests experiments that can be done with a diamond-anvil cell at room temperatures using single crystal ruby as the pressure/stress calibrant. The major advantage of diamond cell studies is the high pressures that can be used under which a sample can be plastically deformed Wenk et al. (2000) recently deformed Fe at ~220 GPa). However, there are major limitations with this approach: homogeneous heating is difficult in a diamond-anvil cell, and almost all previous results were obtained at room temperature. In addition, the strain rate is not well constrained even though deformation is likely to be time-dependent in these tests. Furthermore, the sample space is so small that some important effects such as grain-size sensitivity of strength are difficult to measure. Poirier et al. (1981) and Sotin and Poirier (1990) report on a sapphire cell that enables larger sample volumes than a diamond cell and is more versatile for higher temperature studies. They calibrate the stress field from the characteristics of the loading system and observe the movement of strain markers in the sample. This enables them to define strain rate along with the stress determinations. Deformation-DIA Samples of MgO and polycrystalline tantalum were deformed in the deformationDIA (D-DIA) at the National Synchrotron Light Source (NSLS) in February 2002. This

38

Durham, Weidner, Karato & Wang

marked the first materials test ever above P = 10 GPa in which virtually all relevant rheological independent variables were under full and independent operator control. We provide here a detailed description of this new apparatus. The D-DIA (Wang et al. 2002) is a modification of the cubic anvil device known as the DIA (e.g., Osugi et al. (1964) and Shimomura et al. (1985)) a single-stage, wedged guide block machine. The hydrostatic DIA usually operates to ~15 GPa and ~2000 K. The D-DIA modification gives independent displacement control to one pair of opposing anvils, reducing the cubic symmetry of the DIA to tetragonal, thus allowing high-strain deformation experiments to 15 GPa.

Figure 9. The original hydrostatic DIA apparatus. A single hydraulic ram (large arrows) drives 2 wedge-type guide blocks toward one another. The 6 anvils (truncated pyramids on the guide blocks and wedge blocks; only 5 are visible here) are thus driven toward one another, symmetrically compressing a cubic sample assembly (not shown) at the very center.

The original DIA consists of symmetric upper and lower guide blocks, four wedgeshaped thrust blocks, and six anvils, as indicated in Figure 9. Four of the anvils are mounted on the inside faces of the thrust blocks and the other two are mounted on the inside central faces of the guide blocks, the square fronts of the anvils thus defining a cubic volume at the center of the apparatus. The operation of the DIA can be visualized by recognizing that the eight inner inclined surfaces of the guide blocks define a virtual regular octahedron whose dimension changes with the separation distance between the guide blocks. The six anvils are aligned with the six apices of this virtual octahedron, and parts are machined such that the (fixed) distance from anvil face to associated apex of the virtual octahedron is precisely the same for each anvil. As the guide blocks close or open, all displacements of the anvils are symmetric about the center. The force of a single hydraulic actuator (or ram) applied along the vertical axis in Figure 9 closes the guide blocks and, thus, compresses the cell hydro-statically. Ram forces of 100 T are typically required with various truncation sizes (typically 3 to 6 mm) to reach maximum pressures near 20 GPa (see e.g., Shimomura et al. (1992) and Utsumi et al. (1992)). By the cubic symmetry of the DIA assemby, this force is divided into three equal portions along each of the three orthogonal directions defined by opposing pairs of anvils.

Developments in Deformation Experiments at High Pressure

39

Independent control of one anvil pair in D-DIA is provided by two additional hydraulic actuators—called differential rams—within each guide block (Fig. 10). These rams may be thought of not as parts that are added, but rather as a central portion of each guide block that has been cut free so that it can move vertically. A small hole is introduced into the side of each guide block to allow access for hydraulic fluid to each ram, and seals for the pressurized plenum below each differential ram are provided. The conical shape of the differential rams is rather unimportant; it leaves more steel on the guide blocks and makes them slightly stronger. Note that in the D-DIA modification, the guide blocks not only support the forces confining the thrust blocks, they now have also become thick-wall pressure containers for the hydraulic fluid driving the differential rams. For this reason, the guide blocks of the D-DIA are complete discs rather than crossed steel members as in the original DIA (Fig. 9).

Figure 10. Computer-generated 3/4 section through the D-DIA. Comparing to Figure 9, the guide blocks are now full cylindrical shapes, but the four wedge blocks and six anvils are identical to those in Figure 9. The two differential rams, the large conical pieces within the top and bottom guide blocks, move independently of the guide blocks themselves, and can thus generate a deviatoric state of stress in the assembly at the center (not shown). In actual operation, as the differential rams displace inward toward the center, the guide blocks move apart to allow the four wedge blocks to displace outward. The displacement rates are independently controllable and are usually chosen such that the net volume of the sample assembly, and therefore the confining pressure, remains constant.

The differential rams have a (pressurized) diameter of 89 mm and are therefore capable of generating a force of 125 T at a hydraulic pressure of 0.2 GPa, sufficient to overcome the force of the main ram (up to 33% of the main ram maximum of 300 T) and apply differential force even at the highest confining pressure. Finite element analysis of the D-DIA guide blocks show that the maximum octahedral stress occurs on the inner bore in a “typical” operation where the differential load is 67% of the main ram load (presuming some force is needed to overcome packing friction and sample strength) (Wang et al. 2002). The stress at the base of the inclined surface of the guide block, the

40

Durham, Weidner, Karato & Wang

next highest point of stress concentration in the guide block, is roughly 90% of the stress in the bore. Note that in the original DIA with very large notches cut in the guide blocks, the stress concentration at the base of the inclined surfaces is twice that of the D-DIA under a similar main ram load. The key point of independent control of the differential rams is that deformation can be imposed without the necessity of increasing confining pressure (which has been the long-standing limitation of deformation studies in 6/8 multianvil and diamond-anvil assemblies). In normal operation, the D-DIA sample is brought to run conditions of P and T with differential rams fully withdrawn in the same manner as the hydrostatic DIA. Advancing the differential rams then introduces a non-cubic shape change to the assembly. Furthermore, by simultaneously draining hydraulic fluid from the main ram at an appropriate rate, the four side anvils retract, and the total force of the main ram as well as the volume of the sample cell can be held constant. In the synchrotron X-ray beamline) pressure itself can be monitored and, accordingly, serves as the process variable. The operation at NSLS in February 2002, showed that this procedure gave very satisfactory control of pressure over very large displacements of the differential rams (Fig. 11). The differential rams are driven by high-precision syringe pumps, and their velocities are controllable from approximately 10-7 to 10-2 mm/s. When both rams are driven symmetrically, this translates to a strain rate on a typical 1-mm-length, 1-mm3volume sample of 2 × 10-2 ≥ εÝ ≥ 2 × 10-7 s-1. These rates are much faster than most relevant rates in geology, but as with all experimental rock mechanics work—for which these rates are typical—it is the human time scale and not the geologic time scale that governs the duration of an experiment. As with most experimental studies of rock deformation, appropriate scaling analysis is critical to apply these results to Earth. Differential ram displacements of >1 mm are possible, so strains in pure shear compression can approach 1. Deformation in simple shear, using 45˚-cut pistons (see “Sample Assemblies,” above) is also possible, allowing for much higher strains, at some cost to sample volume. Finally, the D-DIA is capable of extensional, as well compressional, deformation, because the sense of motion of the six anvils can be reversed. Rotational Drickamer apparatus (RDA) Large strain deformation experiments can be conducted in the torsion mode, allowing a detailed study of deformation microstructures as well as rheology. Paterson and Olgaard (2000) described such an apparatus in which a torsion actuator is attached to a gas-medium high-pressure apparatus. This apparatus contains an internal load-cell, making possible a precise measurement of stress. However, the maximum pressure of operation is limited to ~0.5 GPa. Deformation experiments can be conducted at higher pressures by twisting a thin sample between anvils. Bridgman (1935a; 1937) was the first to apply the technique to 5 GPa with an apparatus that consisted of two fixed anvils bearing on a flat, rotating anvil. The design was later modified by Griggs et al. (1960) to remove the intermediate anvil and rotate one of the anvils instead. Similar attempts were made in the late 1960s at room temperatures and at pressures to 7 GPa (Abey and Stromberg 1969; Riecker and Seifert 1964) in which a sample was sheared, and the strength was determined by the measurements of torque on an anvil needed to deform the sample. Yamazaki and Karato (2001) have improved these techniques by modifying the Drickamer-type high-pressure apparatus (Fig. 12). This apparatus allows large rotational shear deformation of a sample at higher pressures and temperatures. A Drickamer apparatus consists of a pair of anvils in a cylinder with a gasket material between the anvils and a thin disk of sample squeezed between the anvils to reach high pressure

Developments in Deformation Experiments at High Pressure

Figure 11. Results plotted vs. run time from a D-DIA experiment on a stacked sample of tantalum (starting length 1 mm) and MgO (starting length 1.5 mm), both polycrystalline. (a) Five traces showing, from top to bottom: measured hydraulic oil pressure in the top and bottom differential rams (scale on right-hand axis); combined centerward displacement of the two rams as measured by displacement transducers outside the sample assembly; shortening of the MgO portion of the sample; and shortening of the Ta component (scale for last three on left-hand axis). Sample length vs. time was measured directly using X-radiography. (b) Detail of the sample shortening, also indicating the strain rates (displacement rates normalized by sample length). The run was conducted by pumping hydraulic fluid to the differential rams at two different constant rates (hence deformation steps (1) and (2)), while draining fluid from the main ram in order to keep pressure at a constant level of 10±1 GPa. Note that shortening rates of the samples were also approximately constant. The steady increase in differential ram pressure during both deformation steps is the result of steadily increasing friction on the gaskets squeezed between the differential and side anvils.

41

42

Durham, Weidner, Karato & Wang

Figure 12. Sketch of a rotational Drickamer apparatus (Yamazaki and Karato 2001).

Figure 13. Sample assembly for a rotational Drickamer apparatus. A thin disk of sample is sandwiched between two anvils. A sample is cut into two pieces and W5%Re and W25%Re foils are inserted vertically; these act both as strain markers and as a thermocouple. The change in shape of these foils can be measured after an experiment or during an experiment through X-ray absorption. A small cylindrical W3%Re foil at the center acts as an electrode for the heater and as one component of the thermocouple with a W25%Re foil.

(Fig. 13). The gasket provides an extra support for the anvils and, as a result, this apparatus yields significantly higher pressures than does a Bridgman apparatus without a gasket (Perez-Albuerne et al. 1964; Prins 1984). Because much of the force is supported by the gasket and the sample itself, the cylinder does not support a large load. Consequently, it is possible to make holes in the cylinder to provide a path for X-rays; the

Developments in Deformation Experiments at High Pressure

43

thermocouple leads can be taken through these holes. The X-ray path feature is important for high-resolution stress and temperature measurements, both of which are critical for rheological studies. The Drickamer apparatus has recently been used for the in situ measurements of static properties using a synchrotron X-ray facility under conditions up to ~35 GPa and ~2000 K (Funamori and Yagi 1993). In the Rotational Drickamer Apparatus (RDA) modification, one of the anvils is fixed with the frame, whereas another is attached to a rotational actuator. A gearbox with an ac-servo motor provides controlled rotation of one of the anvils. A sample is first pressurized (and heated), and then the rotational actuator is started. The sample is twisted between the two anvils, and the geometry of deformation is rotational shear. The strain (and strain rate) increases from zero to the maximum value, which is determined by the motor rotation speed and sample thickness. With the current motor and gear combination, a shear strain rate from ~0 to ~10-3 s-1 can be realized. This design has three advantages. First, the motion of the anvil for deformation is orthogonal to the motion of the anvil for pressurization. Therefore, pressurization and deformation can be clearly separated. Second, unlike axisymmetric deformation, the shape of the sample does not change appreciably during rotational deformation, allowing very large strain deformation and making possible studies of microstructural evolution. Third, because of the radial gradient in strain (strain rate), microstructural evolution and/or rheology at different strain-rates can be investigated in a single run. The diameter of the anvil tip in the current RDA is 4 mm (sample diameter is ~1-2 mm, thickness ~0.2-0.4 mm). With a load of ~30 T, a pressure of ~15 GPa can be generated using tungsten carbide anvils. The maximum pressure is determined primarily by the strength of anvil materials (and the diameter of sample space) and is expected to be more than ~30 GPa when a harder material such as sintered diamond or cubic-BN is used. The rotational Drickamer apparatus constructed at Yale University has been tested at P = 12 GPa, T = 1473 K. A polycrystalline sample of MgO was deformed homogeneously to the maximum shear strain of ~3. Two methods can be employed to estimate the stress magnitude. First, X-ray diffraction techniques can be used for deformation experiments with the RDA in a synchrotron radiation facility in the manner discussed previously. At this writing, such experiments are planned but have not yet been carried out. To adopt the diffraction technique for rotation geometry, a conical shape window is made in a cylinder to collect diffracted X-rays for a range of angles (Fig. 14b). One complication with this apparatus is the radial variation of stress. This is inevitable in the torsion test (e.g., Paterson and Olgaard 2000); Fig. 14a). In measuring the stress magnitude by X-ray diffraction, one must consider the spatial resolution. Important factors that control the spatial resolution are the cross-sectional area of the X-ray beam and the 2θ value used for diffraction (with energy dispersive mode). A smaller beam size and a larger 2θ are preferred for a better spatial resolution of stress measurements. However, a smaller beam size results in a weaker signal. Also the 2θ values are limited by the energy of X-ray and the d-spacings as well as the geometry of the apparatus (we will use ~10-50 × 10-50 μm2 beam [slit] size and 2θ ≈ 8-10°). To allow for control of 2θ, the apparatus is set on a mobile stage (with a remote control through a stepping motor) whose angle and position with respect to the beam line can be adjusted to optimize the conditions. The resolution of the stress measurement is also dependent on the detector. We anticipate that a spatial resolution of 100-200 μm and the stress resolution better than 0.1 GPa will be obtained by this technique. The use of a hollow specimen could help solve the problem of stress heterogeneity (Fig. 14d). We note that this technique not only allows us to estimate the stress magnitude, but also provides information as to the grain-scale stress (strain) distribution in a polycrystalline material (either homogeneous stress or homogeneous strain (Funamori et

44

Durham, Weidner, Karato & Wang

..

n

Figure 14. (A) Stress distribution in a sample in a torsion test (n is stress exponent, γ = Aτ ). (B) X-ray diffraction geometry. Stress magnitude can be estimated from the difference in d-spacing collected from two different diffracted beams through two windows that are cut at 45o with respect to the vertical axis. (C) A top view of a sample and the X-ray path. In order to determine the stress, X-ray diffraction must occur in a region of a sample where the shear direction is normal to the X-ray beam. (D) The geometry of a hollow sample, where stress heterogeneity is less.

al. 1994). Such information is useful in understanding the physics of deformation of polycrystals (Kocks 1970). Another method is to use dislocation density as a stress indicator. This technique can be used when the deformation mechanism is dislocation creep and when stress has been held at a constant value for at least a few percent strain. The method does not work well if stress is inhomogeneous. We have recently used this technique for olivine and are able to estimate the stress magnitude with an uncertainty of ~10-15% (Jung and Karato 2001; Karato and Jung 2002 in press). The reliability of such measurements for high-pressure phases is not as high as that for olivine because the calibration curve for dislocation density vs. stress relationship is not available. However, the relation between dislocation density and stress is nearly universal,

ρ = α b − 2 (σ / μ ) 2 (ρ: dislocation density, α: a constant of order unity, b: the length of the Burgers vector, σ: stress, μ: shear modulus (e.g., Kohlstedt and Weathers 1980) We consider that such measurements should provide at least semi-quantitative data on the stress magnitude. We will perform both single-layer deformation experiments and two-layer deformation experiments. A single-layer deformation experiment is simpler to interpret, but the big advantage of a two-layer design is that one can determine the relative strength of two samples very accurately. Our experience with this apparatus showed that we can deform our samples to strains up to at least γ ≈ 10 at ~15 GPa. Such a large strain is essential for the study of deformation fabric. SUMMARY AND PERSPECTIVES We have summarized some of the recent developments in quantitative characterization of stress and strain and the controlled generation of deviatoric stresses under high pressure (and temperature). These developments will allow us to explore

Developments in Deformation Experiments at High Pressure

45

rheological behavior of Earth materials under much deeper conditions than heretofore possible. Such studies will, for the first time, provide the critical data sets necessary for understanding whole mantle dynamics through modeling and/or through the analyses of observations such as seismic anisotropy. Further technical developments need to be made to make these techniques robust. First, stress measurements using different techniques must be benchmarked. Three techniques are currently used to measure stress under high pressure: (1) Measurement of force using an external load-cell with a low-friction sample design (Tingle et al. 1993), (2) synchrotron X-ray measurement of orientation dependence of d spacing (Singh et al. 1998), and (3) inference from dislocation densities (Karato and Jung 2002 in press). Below ~4 GPa, all of these techniques can be employed to determine the strength of a standard materials such as olivine. A comparison of results of these measurements will provide a measure of reliability of each technique. Second, a number of issues still remain regarding the sample assembly. They include a proper control of thermodynamic environment, particularly water fugacity (or water content) and oxygen fugacity. Third, in almost all of these techniques under high pressure, it is almost inevitable to have deformation during pressurization. An experimental procedure (e.g., annealing; a very low strength medium surrounding the sample) must be established to minimize the effects of initial stage deformation. It must also be emphasized that the sample preparation and characterization are as critical as careful mechanical tests. Samples with negligible porosity must be used and grain size and water content must be measured both before and after each experiment (control of other thermodynamic variables such as oxygen fugacity is also important; see Rubie et al. (1993)). In almost all laboratory deformation experiments, conditions are often close to the boundary between dislocation and diffusion creep (Karato et al. 1998), and, therefore, precise measurements of grain size are needed to interpret the data. Also, water content must be measured both before and after each experiment. Water is known to have significant effects on rheology, but its content in samples is difficult to control during high-pressure experiments. Using the results of mechanical tests and sample characterization, dominant microscopic mechanisms of deformation must be identified. A meaningful comparison of strength of different materials (or for the same material for different conditions) can only be made when the deformation mechanism is the same, and the comparison must be made based on an appropriate scaling. For instance, if the dominant mechanism of deformation is diffusion creep, then the data must be compared by normalizing with respect to the grain size. With the use of these new techniques (combined with careful characterization of samples and the analyses of deformation mechanisms), we will be able to cast first light on the rheology of the more than 90% of Earth’s mantle that has not been accessible by quantitative studies. Results of such studies will be critical to our better understanding of dynamics and evolution of this planet. ACKNOWLEDGMENTS We are most grateful to H.W. Green, II, for reviewing the manuscript. Funding was provided by the National Science Foundation under award EAR0135551. Work by WBD performed under the auspices of the U.S. Department of Energy by the Lawrence Livermore National Laboratory under contract W-7405-ENG-48. Additional portions of this work were performed at GeoSoilEnviroCARS (GSECARS), Sector 13, Advanced Photon Source at Argonne National Laboratory. GSECARS is supported by the National Science Foundation–Earth Sciences, Department of Energy– Geosciences, W.M. Keck Foundation, and the U.S. Department of Agriculture.

46

Durham, Weidner, Karato & Wang

Use of the Advanced Photon Source was supported by the U.S. Department of Energy, Basic Energy Sciences, Office of Energy Research, under Contract No. W-31109-Eng-38. REFERENCES Abey AE, Stromberg HD (1969) 70 kilobar shear apparatus. Rev Sci Instrum 40:557 Borch RS, Green HW II (1989) Deformation of peridotite at high pressure in a new molten cell: Comparison of traditional and homologous temperature treatments. Phys Earth Planet Inter 55: 269-276 Boyd FR, England JL (1960) Apparatus for phase equilibrium measurements at pressures up to 50 kilobars and temperatures up to 1750˚C. J Geophys Res 65:741-748 Bridgman PW (1914) The technique of high pressure experimenting. Proc Am Acad Arts Sci 49:627-643 Bridgman PW (1935a) Effects of high shearing stress combined with high hydrostatic pressure. Phys Rev 48:825-847 Bridgman PW (1935b) Polymorphism, principally of the elements, up to 50,000 kg/cm2. Phys Rev 48: 893-906 Bridgman PW (1937) Shearing phenomena at high pressures, particularly in inorganic compounds. Proc Am Acad Arts Sci 72:45-136 Bridgman PW (1942) Pressure-volume relations for seventeen elements to 100,000 kg/cm2. Proc Am Acad Arts Sci 74:425-440 Bridgman PW (1948) The compression of 39 substances to 100,000 kg/cm2. Proc Am Acad Arts Sci 76: 55-70 Bridgman PW (1952a) The resistance of 72 elements, alloys, and compounds to 100,000 kg/cm2. Proc Am Acad Arts Sci 81:65-NN Bridgman PW (1952b) The Physics of High Pressure. G. Bell and Sons, London Bridgman PW (1952c) Studies in Large Plastic Flow and Fracture with Special Emphasis on the Effects of Hydrostatic Pressure. McGraw-Hill, New York Burnley PC, Green HW II, Prior D (1991) Faulting associated with the olivine to spinel transformation in Mg2GeO4 and its implications for deep-focus earthquakes. J Geophys Res 96:425-443 Bussod GY, Katsura T, Rubie DC (1993) The large volume multi-anvil press as a high P-T deformation apparatus. Pure Appl Geophys 141:579-599 Carter N.L, Christie J.M, Griggs D. T (1964) Experimental deformation and recrystallization of quartz. J Geol 72:687-733 Chai M, Brown JM (1996) Effects of static non-hydrostatic stress on the R lines of ruby single crystals. Geophys Res Lett 23:3539-3542 Chen J, Inoue T, Weidner DJ, Wu Y, Vaughan MT (1998) Strength and water weakening of mantle minerals, olivine, wadsleyite and ringwoodite. Geophys Res Lett 25:575-578 Cordier P, Rubie DC (2001) Plastic deformation of minerals under extreme pressure using a multi-anvil apparatus. Mater Sci Eng A309-310:38-43 Davies GF (1995) Penetration of plates and plumes through the mantle transition zone. Earth Planet Sci Lett 133, 507-516 Durham WB, Rubie DC (1998) Can the multianvil apparatus really be used for high-pressure deformation experiments? In Properties of Earth and Planetary Materials at High Pressure and Temperature, Geophysical Monograph 101. M Manghnani, Y Yagi (eds) p 63-70. American Geophysical Union Eremets M. I (1996) High Pressure Experimental Methods. Oxford University Press Fujimura A (1989) Preferred orientation of mantle minerals. In Rheology of Solids and of the Earth. S-i Karato, M Toriumi (eds) p 263-283. Oxford University Press Fujimura A, Endo S, Kato M, Kumazawa M (1981) Preferred orientation of β−Mn2GeO4. In Programme and Abstracts. Japan Seismological Society, p 185 Fukao Y, Obayashi M, Inoue H, Nenbai M (1992) Subducting albs stagnant in the mantle transition zone. J Geophys Res 97:4809-4822 Funamori N, Yagi T (1993) High pressure and high temperature in situ X-ray observation of MgSiO3 perovskite under lower mantle conditions. Geophys Res Lett 20:387-390 Funamori N, Yagi T, Uchida T (1994) Deviatoric stress measurement under uniaxial compression by a powder X-ray diffraction method. J Appl Phys 75:4327-4331 Getting IC (1998) New determination of the bismuth I-II equilibrium pressure: A proposed modification to the practical pressure scale. Metrologia 35:119-132 Getting IC, Spetzler HA (1993) Gas-charged piston-cylinder apparatus for pressures to 4 GPa. In HighPressure Science and Technology. SC Schmidt, JW Shaner, GA Samara, M Ross (eds) p 1581-1584. American Institute of Physics

Developments in Deformation Experiments at High Pressure

47

Green HW II (2001) Earthquakes at depth and their enabling mineral reactions. Eleventh Annual V.M. Goldschmidt Conf (abstr) Green HW II, Borch RS (1989) A new molten salt cell for precision stress measurement at high pressure. European J Mineral 1:213-219 Green HW II, Houston H (1995) The mechanics of deep earthquakes. Ann Rev Earth Planet Sci 23:169-213 Green HW II, Scholz CH, Tingle TN, Young TE, Koczynski TA (1992) Acoustic emissions produced by anticrack faulting during the olivine-spinel transformation. Geophys Res Lett 19:789-792 Green HW II, Young TE, Walker D, Scholz CH (1990) Anticrack-associated faulting at very high pressure in natural olivine. Nature 348:720-722 Griggs DT (1936) Deformation of rocks under high confining pressure. J Geol 44:541-577 Griggs DT (1967) Hydrolytic weakening of quartz and other silicates. Geophys J R Astron Soc 14:19-31 Griggs DT, Turner FJ, Heard HC (1960) Deformation of rocks at 500˚ to 800˚C. In Rock Deformation, Geol Soc Am Memoir 79. DT Griggs, JW Handin (eds) p 39-104. Geol Soc Am, Boulder, Colorado Hager BH (1984) Subducted slabs and the geoid: constraints on mantle rheology and flow. J Geophys Res 89:6003-6015 Hall HT (1958) Some high-pressure, high-temperature apparatus design considerations: Equipment for use at 100,000 atmospheres and 3000˚C. Rev Sci Instrum 29:267-275 Handin JW (1953) An application of high pressure in geophysics: experimental rock deformation. Am Soc Mech Eng Trans 75:315-325 Heard HC, Carter NL (1968) Experimentally induced "natural" intergranular flow in quartz and quartzite. Am J Sci 266:1-42 Hill R (1950) The Mathematical Theory of Plasticity. Oxford University Press Hirth G, Kohlstedt DL (1996) Water in the oceanic upper mantle; implications for rheology, melt extraction and the evolution of the lithosphere. Earth Planet Sci Lett 144:93-108 Hobbs BE, Ord A (1988) Plastic instabilities: implications for the origin of intermediate and deep focus earthquakes. J Geophys Res 89:10521-10540 Honda S, Yuen DA, Balachandar S, Reuteler D (1993) Three-dimensional instabilities of mantle convection with multiple phase transitions. Science 259:1308-1311 Houck JC, Hutton UO (1963) Correlation of factors influencing the pressures generated in multi-anvil devices. In High-Pressure Measurement. AA Giardini, EC Lloyd (eds). Butterworths, London Jaeger JC, Cook NGW (1976) Fundamentals of Rock Mechanics, 2nd edition. Chapman and Hall Jiao W, Siliver PG, Fei Y, Prewitt CT (2000) Do deep earthquakes occur on preexisting weak zones? An examination of the Tonga subduction zone. J Geophys Res 105:28125-28138 Jung H, Karato S (2001) Effect of water on the size of dynamically recrystallized grains in olivine. J Struct Geol 23:1337-1344 Karato S-i (1998) Micro-physics of post-glacial rebound. In Ice Age Dynamics. P Wu (ed) p 351-364. TTP (Trans Tech Publications), Zürich Karato S-i (1998a) Effects of pressure on plastic deformation of polycrystalline solids: some geological applications. In High Pressure Research in Materials Sciences. PY Yu (ed) p 3-14. Materials Research Society, Warrendale, Pennsylvania Karato S-i, Dupas-Bruzek C, Rubie DC (1998) Plastic deformation of silicate spinel under the transition zone conditions of the Earth. Nature 395:266-269 Karato S-i, Jung H (2002) Effects of pressure on high-temperature dislocation creep in olivine polycrystals. Phil Mag (in press) Karato S-i, Riedel MR, Yuen DA (2001) Rheological structure and deformation of subducted slabs in the mantle transition zone: implications for mantle circulation and deep earthquakes. Phys Earth Planet Inter 127:83-108 Karato S-i, Rubie DC (1997) Toward experimental study of plastic deformation under deep mantle conditions: a new multianvil sample assembly for deformation studies under high pressures and temperatures. J Geophys Res 102:20111-20122 Karato S-i, Wu P (1993) Rheology of the upper mantle; a synthesis. Science 260:771-778 Kawai N, Endo S (1970) The generation of ultrahigh hydrostatic pressure by a split sphere apparatus. Rev Sci Instrum 41:1178-1181 Kellogg LH, Hager BH, van der Hilst RD (1999) Compositional stratification in the deep mantle. Science 283:1881-1884 King SD (1995) Radial models of mantle viscosity: results from a generic algorithm. Geophys J Intl 122:725-734 Kinsland GL, Bassett W (1977) Strength of MgO and NaCl polycrystals to confining pressures of 250 kbar at 25 C. J Appl Phys 48:978-985 Kirby SH, Stein S, Okal EA, Rubie DC (1996) Metastable mantle phase transformations and deep earthquakes in subducting oceanic lithosphere. Rev Geophys 34:261-306

48

Durham, Weidner, Karato & Wang

Kocks UF (1970) The relation between polycrystal deformation and single-crystal deformation. Metall Trans 1:1121-1143 Kohlstedt DL, Chopra PN (1994) Influence of basaltic melt on the creep of polycrystalline olivine under hydrous conditions. In Magmatic Systems. Intl Geophysics Series, Vol. 57. MP Ryan (ed) p 37-53 Kohlstedt DL, Weathers MS (1980) Deformation-induced microstructures, paleopiezometers, and differential stresses in deeply eroded fault zones. J Geophys Res 85:6269-6285 Liebermann RC, Wang Y (1992) Characterization of sample environment in a uniaxial split-sphere apparatus. In High-Pressure Research: Application to Earth and Planetary Sciences. Y Syono, MH Manghnani (eds) p 19-31. American Geophysical Union, Washington, DC Lloyd EC, Hutton UO, Johnson DP (1959) Compact multi-anvil wedge-type high pressure apparatus. J Res Nat Bur Stand 59:63C Masters G, Laske G, Bolton H, Dziewonski AM (2000) The relative behavior of shear velocity, bulk sound speed, and compressional velocity in the mantle: Implications for chemical and thermal structure. In Earth's Deep Interior. S Karato, AM Forte, RC Liebermann, G Masters, L Stixrude (eds) p 63-87. American Geophysical Union, Washington, DC Meade C, Jeanloz R (1988a) The yield strength of B1 sand B2 phases of NaCl. J Geophys Res 93: 3270-3274 Meade C, Jeanloz R (1988b) The yield strength of MgO to 40 GPa. J Geophys Res 93:3261-3269 Meade C, Jeanloz R (1990) The strength of mantle silicates at high pressures and room temperature: implications for the viscosity of the mantle. Nature 348:533-535 Mei S, Kohlstedt DL (2000a) Influence of water on plastic deformation of olivine aggregates, 1. Diffusion creep regime. J Geophys Res 105:21457-21469 Mei S, Kohlstedt DL (2000b) Influence of water on plastic deformation of olivine aggregates, 2. Dislocation creep regime. J Geophys Res 105:21471-21481 Mitrovioca JX, Forte AM (1997) Radial profile of mantle viscosity: results from the joint inversion of convection and postglacial rebound observables. J Geophys Res 102:2751-2769 Mitrovioca JX, Peltier WR (1991) A complete formalism for the inversion of postglacial rebound data: resolving power analysis. Geophys. J Intl 104:267-288 Ogawa M (1987) Shear instability in a viscoelastic material as the cause for deep earthquakes. J Geophys. Res 92:13801-13810 Onodera A (1987) Octahedral-anvil high-pressure devices. High Temp-High Press 19:579-609 Osugi J, Shimizu K, Inoue T, Yasunami K (1964) A compact cubic anvil high pressure apparatus. Rev Phys Chem Japan. 34:1-6 Paterson M. S (1970) A high-pressure, high-temperature apparatus for rock deformation. Intl J Rock Mech Min Sci 7:517-526 Paterson MS (1978) Experimental Rock Deformation: The Brittle Field. Springer-Verlag, Berlin Paterson MS (1990) Rock deformation experimentation. In The Brittle-Ductile Transition in Rocks: The Heard Volume. AG Duba, WB Durham, JW Handin, HF Wang (eds) p 187-194. American Geophysical Union, Washington, DC Paterson MS, Olgaard DL (2000) Rock deformation tests to large shear strains in torsion. J Struct Geol 22:1341-1358 Peltier WR (1998) Postglacial variation in the level of the sea: implications for climate dynamics and solidEarth geophysics. Rev Geophys 36:603-689 Perez-Albuerne EA, Forgsgren KF, Drickamer HG (1964) Apparatus for X-ray measurements at very high pressure. Rev .Sci Instrum 35, 29-33 Poirier J-P, Sotin C, Peyronneau J (1981) Viscosity of high-pressure ice VI and evolution and dynamics of Ganymede. Nature 292:225-227 Post RL (1977) High-temperature creep of Mt. Burnett dunite. Tectonophysics 42:75-110 Prins JF (1984) A semiempirical description of pressure generation between Bridgman anvils. High Temp – High Press 16:657-664 Raleigh CB, Paterson MS (1965) Experimental deformation of serpentinite and its tectonic implications. J Geophys Res 70:3965-3985 Riecker RE, Seifert KE (1964) Olivine shear strength at high pressure and room temperature. Geol Soc Am Bull 75:571-574 Rubey WW (1972) Foreword. In Flow and Fracture of Rocks. Geophys Monogr 16:ix-x. HC Heard, IY Borg, NL Carter, CB Raleigh (eds) p. American Geophysical Union, Washington, DC Rubie DC, Karato S-i, Yan H, O'Neill HSC (1993) Low differential stress and controlled chemical environment in multianvil high-pressure experiments. Phys Chem Minerals 20:315-322 Shimomura O, Utsumi W, Taniguchi T, Kikegawa T, Nagashima T (1992) A new high pressure and high temperature apparatus with sintered diamond anvils for synchrotron radiation use. In High-pressure Research: Application to Earth and Planetary Sciences. Y Syono, MH Manghnani (eds) p 3-11. Terra Scientific Publishing, Tokyo

Developments in Deformation Experiments at High Pressure

49

Shimomura O, Yamaoka S, Yagi T, Wakatsuki M, Tsuji K, Kawamura H, Hamaya N, Fukuoga O, Aoki K, Akimoto S (1985) Multi-anvil type X-ray system for synchrotron radiation. In Solid State Physics Under Pressure. S Minomura (ed), p 351-356. Terra Scientific Publishing, Tokyo Singh AK, Balasingh C, Mao H-k, Hemley R, Shu J (1998) Analysis of lattice strains measured under nonhydrostatic pressure. J Appl Phys 83:7567-7578 Sotin C, Poirier J-P (1990) The sapphire anvil cell as a creep apparatus. In The Brittle-Ductile Transition in Rocks: The Heard Volume, Vol. Geophys. Monogr. No. 56. AG Duba, WB Durham, JW Handin, HF Wang (eds) p 219-223. American Geophysical Union, Washington, DC Sung CM, Goetze C, Mao HK (1977) Pressure distribution in the diamond anvil press and the shear strength of fayalite. Rev Sci Instrum 48:1386-1391 Tackley PJ, Stevenson DJ, Glatzmaier GA, Schubert G (1993) Effects of endothermic phase transition at 670 km depth in a spherical model of mantle convection in the Earth's mantle. Nature 361:699-704 Thoraval C, Richards MA (1997) The geoid constraint in global geodynamics: viscosity structure, mantle heterogeneity models and boundary conditions. Geophys J Intl 131:1-8 Tingle TN, Green HW II, Young TE, Koczynski TA (1993) Improvements to Griggs-type apparatus for mechanical testing at high pressures and temperatures. Pure Appl Geophys 141:523-543 Tullis TE, Tullis J (1986) Experimental rock deformation techniques. In Mineral and Rock Deformation: Laboratory Studies: The Paterson Volume. BE Hobbs, HC Heard (eds) p 297-324. American Geophysical Union, Washington, DC Turner FJ, Griggs DT, Clark RH, Dixon RH (1956) Deformation of Yule marble. Part VII: Development of oriented fabrics at 300˚-500˚C. Geol Soc Am Bull 67:1259-1294 Utsumi W, Yagi T, Leinenweber K, Shimomura O, Taniguchi T (1992) High-pressure and hightemperature generation using sintered diamond anvils. In High-pressure Research: Application to Earth and Planetary Sciences. Y Syono, MH Manghnani (eds), p 37-42. American Geophysical Union, Washington, DC van der Hilst RD, Engdahl R, Spakman W, Nolet G (1991) Tomographic imaging of subducted lithosphere below northwest Pacific island arcs. Nature 353:37-43 van der Hilst RD, Karason H (1999) Compositional heterogeneity in the bottom 1000 kilometers of Earth's mantle: toward a hybrid convection model. Science 283:1885-1888 van der Hilst RD, Widiyantoro RDS, Engdahl ER (1997) Evidence for deep mantle circulation from global tomography. Nature 386:578-584 Vaughan M, Chen J, Li L, Weidner DJ, Li B (2000) Use of X-ray imaging techniques at high pressure and temperature for strain measurements. In Science and Technology of High Pressure. Proceedings of AIRAPT-17. MH Manghnani, WJ Nellis, MF Nicol (eds) p 1097-1098. Universities Press Walker D (1991) Lubrication, gasketing, and precision in multianvil experiments. Am Mineral 76: 1092-1100 Walker D, Carpenter MA, Hitch CM (1990) Some simplifications to multianvil devices for high-pressure experiments. Am Mineral 75:1020-1028 Wang Y, Durham W, Getting I (2002) D-DIA: A new apparatus for high-temperature deformation at pressures up to 15 GPa. Rev Sci Instrum (submitted) Weidner DJ (1998) Rheological studies at high pressure. Rev Mineral 39:493-524 Weidner DJ, Chen J, Xu Y, Wu Y, Vaughan MT, Li L (2001) Subduction zone rheology. Phys Earth Planet Inter 127:67-81 Weidner DJ, Vaughan MT, Ko J, Wang Y, Liu X, Yeganeh-Haeri A, Pacalo RE, Zhao Y (1992) Characterization of stress, pressure, and temperature in SAM85, A DIA type high pressure apparatus. In High-Pressure Research: Application to Earth and Planetary Sciences. Geophys Monogr 67:13-17. Y Syono, MH Manghnani (eds) Terra Scientific Publishing Company, Tokyo, and American Geophysical Union, Washington, DC Weidner DJ, Wang Y, Chen G, Ando J, Vaughan MT (1998) Rheology measurements at high pressure and temperature. In Properties of Earth and Planetary Materials at High Pressure and Temperature. Geophys Monogr 101:473-482. M Manghnani, Y Yagi (eds) American Geophysical Union, Washington, DC Weidner DJ, Wang Y, Vaughan MT (1994) Strength of diamond. Science 266:419-422 Wenk H-R, Matthies S, Hemley RJ, Mao HK, Shu J (2000) The plastic deformation of iron at pressure of the Earth's inner core. Nature 405:1044-1047 Wu TC, Bassett WA (1993) Deviatoric stress in a diamond anvil cell using synchrotron radiation with two diffraction geometries. Pure Appl Geophys 141:509-519 Yamazaki D, Karato S (2001) High pressure rotational deformation apparatus to 15 GPa. Rev Sci Instrum 72:4207-4211

3

Deformation of Granitic Rocks: Experimental Studies and Natural Examples Jan Tullis Department of Geological Sciences Brown University Providence, Rhode Island 02912

INTRODUCTION Crustal deformation is inherently complex because the crust is inhomogeneous and anisotropic on a wide variety of scales. Advances in our understanding have come from a combination of field observations, laboratory experiments and theoretical models. What necessarily links all these investigations is a focus on process. Crystal plasticity (dislocation creep) is one of several grain-scale deformation processes or mechanisms; each process is dominant over a certain range of conditions in the crust, has a particular form of the flow law, and produces a characteristic set of microstructures. Characterizing the deformation mechanisms for different minerals, over a range of conditions, allows one to link the mechanical behavior or flow law associated with that process to the microstructures it produces. The information on microstructures can be used to help interpret the thermomechanical history of naturally deformed rocks, and flow laws can be used to create more realistic models predicting crustal behavior in different tectonic settings. Iterations among field, theoretical, and experimental approaches are critical for making progress in our understanding. The deformation process of crystal plasticity is the main focus of this volume. Crystal plastic deformation of the crust is in many ways more complex than that of the mantle, because most crustal rocks are polyphase aggregates, and the strengths of common crustal silicates (such as quartz, feldspar, and mica) are very different at a given set of conditions. Yield strength envelopes for the crust, as shown schematically in Figure 1, are commonly plotted using the dislocation creep flow law for quartz, which inherently assumes that quartz is the controlling phase and undergoes steady state flow. But is the flow law for dislocation creep of quartz a good approximation for the deformation of a

Figure 1. Highly schematic depiction of maximum rock strength as a function of depth for the continental lithosphere. The transition from frictional sliding to crystal plasticity is blunted to indicate the operation of other mechanisms such as cataclastic flow and pressure solution. There may be multiple brittle-viscous transitions, depending on compositional layering in the crust, and the depth of the peak strength will depend on the local geotherm, rock and fluid composition, and previous deformation history.

1529-6466/00/0051-0003$05.00

52

Tullis

granite, let alone a crustal ‘plum pudding’ of regions and layers of different lithologies? The crust also tends to be anisotropic on various scales, due to sedimentary layering, metamorphic foliation, igneous intrusions, lattice preferred orientations (LPOs), and ductile shear zones with reduced grain size. How does this anisotropy evolve with time (strain) and how does it influence crustal strength and strain partitioning? The chemical environment is also important; trace amounts of water profoundly affect the dislocation creep strength of various minerals. Is it possible that ‘dry’ granulite grade rocks might be stronger than the colder but wetter rocks above them, creating a weak layer in the middle lower crust? In this contribution I will focus on crystal plasticity in granitic rocks of the continental crust, emphasizing some of the progress that has been made in the recent past from experimental work, with a particular focus on microstructures, and indicating the similarities and differences that are observed in nature. I will begin with brief summaries of our understanding of crystal plasticity in monomineralic aggregates of quartz and of feldspar, then briefly illustrate the similarities and differences in deformation behavior of these phases in polyphase aggregates. In order to induce crystal plasticity in silicates at experimental strain rates about 5 to 8 orders of magnitude faster than natural rates, it is necessary to trade temperature for time. Early attempts to induce ductile deformation in quartz were unsuccessful; although quartz was obviously weak even at low metamorphic grade in nature, it was strong and brittle at the conditions achievable in the early generations of deformation apparatus, and some workers postulated it must ‘flow’ by penetrative fracturing and rehealing. It was Griggs’ development of apparatus capable of operation at higher confining pressures (to inhibit cracking and, as was later realized, to provide high water fugacity) and temperatures (to increase diffusion rates), plus his and Blacic’s discovery of the importance of trace amounts of water, that finally allowed crystal plasticity to be induced experimentally in single crystals (Griggs and Blacic 1964) and ‘mylonites’ to be produced from quartzites (e.g., Carter et al. 1964; Tullis et al. 1973). How can we be sure that the crystal plasticity induced in experimental samples is the same process that occurs in nature? Primarily by verifying that the deformation microstructures (grain size, shape, internal defects and LPO) are the same, but also by extrapolating the experimental flow laws to natural conditions and seeing if the predicted stresses appear reasonable. Experimental results on crystal plasticity have proved very useful in understanding natural crustal deformation, but field observations have also shown the importance of other processes, such as a switch to grain-size sensitive creep induced by syntectonic reactions. These applications and interpretations are discussed in the final section, together with an assessment of our current understanding and remaining questions concerning the strength of the crust. CRYSTAL PLASTICITY PROCESSES For the grain-scale deformation mechanism of crystal plasticity, strain (change of shape) occurs by propagation of line defects or dislocations through crystals (for a more detailed review see Poirier 1985 or Green 1992). These dislocations glide on specific crystal planes, generally those within which bonds are stronger and across which bonds are fewer and/or weaker, and in specific crystal directions, generally those with shorter unit cell parameters. The combination of a slip plane and direction is termed a slip system and may be specified as basal or (001) [110], for example. Glide of dislocations occurs by expansion of dislocation loops, with edge dislocation components (extra half planes) at the leading and trailing portions propagating in the shear direction, and screw components on the sides propagating normal to the shear direction (Fig. 2a). This

Deformation of Granitic Rocks

53

incremental glide accomplishes ‘card deck shear’, with an individual offset equal to the Burgers vector, normally one unit cell length. In contrast to olivine, individual dislocations in quartz and feldspar cannot be observed optically, but arrays of geometrically necessary dislocations produce optical microstructures such as undulatory extinction (Fig. 2b) and low angle tilt or subgrain boundaries (Fig. 2c).

Figure 2. Schematic illustrations of dislocations. (a) Idealization of half of a dislocation loop consisting of pure edge and screw components (heavy line). Light solid and dashed lines represent crystal lattice planes normal to the glide plane of the dislocation; b indicates the Burgers vector; T symbols are edge dislocations (extra half planes of atoms). An applied shear stress (heavy arrows) causes expansion of the loop (small arrows), increasing the slipped area (stippled). [Used by permission of Cambridge University Press. From Poirier (1985) Creep of Crystals, Fig. 2.4, p. 52.] (b) Schematic representation of undulatory extinction, with distributed edge dislocations of both signs ‘wedging’ the crystal lattice. (c) Idealized low angle (tilt) subgrain boundary, consisting of an array of edge dislocations. Both glide and climb are required to produce such a boundary.

The stress required for dislocation glide is much lower than that required for fracture because only a few bonds are broken at a time. Experiments on oriented single crystals can be used to determine the critical resolved shear stress (CRSS) for slip on particular crystallographic planes and directions; these CRSS values may change with deformation conditions such as temperature and water fugacity. Most silicates are anisotropic so that one slip plane is significantly easier than others; thus even in pure monophase aggregates, grains in some orientations will be relatively ‘hard’ and others will be ‘soft’. Dislocation creep refers to steady state flow, which requires a recovery process in addition to the strain-producing glide process. Glide alone tends to result in work hardening, from the tangling of dislocations gliding on intersecting planes or from obstacles such as impurities on the glide plane. Continued strain at constant stress requires an accommodation or recovery process that continually reduces the dislocation density by annihilation and/or rearranges the dislocations into low energy configurations. There are several different recovery processes. At relatively high temperatures where volume diffusion is rapid, climb of edge dislocations allows them to maintain low energy configurations (e.g., Fig. 2c), and to get past obstacles and continue gliding at constant

54

Tullis

stress. At these conditions the creep process may be termed climb-accommodated dislocation creep. Cross slip of screw dislocations also may contribute to recovery. A different recovery process occurs at low temperatures where volume diffusion is slow and climb is limited. At such conditions, single crystals work harden up to fracture, but in aggregates the grain boundaries allow bulging recrystallization which can serve to soften or recover the structure. This process is driven by high dislocation density contrasts across local grain boundary segments, and results in the formation of very small and initially dislocation-free recrystallized grains (e.g., Bailey and Hirsch 1962). The recrystallized grains achieve a dynamic steady state of continual work hardening and recovery; this process may be termed recrystallization-accommodated dislocation creep (Hirth and Tullis 1992). At all conditions, dynamic recrystallization may relax grain boundary stress and strain mismatches and thus reduce the number of active slip systems from the five that are required for a completely homogeneous strain. The operation of dislocation creep produces characteristic microstructures, on the optical to TEM scale, which can potentially provide a variety of useful information about the thermomechanical history of naturally deformed rocks. First, there are several different processes of dynamic recrystallization, dominant at different temperatures (flow stresses), which produce distinct microstructures (e.g., Urai et al. 1986; Hirth and Tullis 1992). Thus optical characterization of the size, shape and internal strain energy of porphyroclasts and recrystallized grains can provide useful information about the deformation conditions. Second, transmission electron microscope (TEM) characterization of the type, density and arrangement of dislocations can be used to provide information about dominant slip systems from natural samples, although details of the dislocation microstructure may have changed following the deformation of interest. Third, materials research shows a relation between the magnitude of the steady-state flow stress and the free dislocation density, the subgrain size, and the recrystallized grain size (Takeuchi and Argon 1976). Once experimental calibrations are done for different recrystallization mechanisms in different minerals, such piezometers may allow constraints to be put on the magnitude of the flow stress from different depths and tectonic settings. Fourth, dislocation creep produces LPOs, because slip on specific planes tends to rotate the crystal with respect to the instantaneous stretching axes of bulk flow. Different patterns, reflecting different dominant slip systems as well as different strain geometries, can be used to indicate different deformation conditions. Dislocation creep is a thermally activated process, and the most commonly used flow law is a power law relation between strain rate ε and differential stress σ1 – σ3 ε = A(σ1-σ3)n exp(-Q/RT)

(1)

where A is a material parameter, the stress exponent n is typically 3 to 5, Q is the activation energy, R is the gas constant, and T is absolute temperature. For dislocation creep there is no dependence on grain size; thus the strength of materials deforming by this process can be significantly different from that for materials deforming by grain size sensitive creep processes (grain boundary sliding and diffusion creep). A chemical activity or fugacity term typically needs to be included in the flow law to account for the effect of the chemical environment, e.g., ε ∝ (fH2O)m (Kohlstedt et al. 1995). Flow law determinations are usually made using gas apparatus (operated at ≤300 MPa), due to their superior stress sensitivity and capability to operate in constant stress mode, although the development of the molten salt cell (MSC) for high pressure Griggs-type apparatus (Green and Borch 1989) has greatly improved the quality of mechanical data achievable at higher pressures (500 MPa to 5 GPa). It is important to note that a flow law can only be extrapolated outside of experimental conditions if it has been determined for a single process which has achieved steady state flow. Flow law determinations are typically done

Deformation of Granitic Rocks

55

using a series of small strain steps (1-5%) and small total strains (10-20%); in such cases it is important to verify that microstructural re-equilibration has occurred. Obviously a major question concerning extrapolation of experimental flow laws is whether steady state flow by the same (single) process has occurred in the natural area of interest. The following sections present brief summaries of the processes, microstructures and mechanical behavior or flow laws (where available) for dislocation creep in pure quartz aggregates and pure feldspar aggregates, primarily based on laboratory experiments but including comparisons with naturally deformed examples. These sections are followed by accounts of some of the complications observed to occur in the experimental deformation of polyphase aggregates of quartz, feldspar and mica, and a comparison with granitic rocks naturally deformed over a wide range of conditions. The last section addresses our current understanding and remaining questions concerning the role of crystal plasticity in crustal deformation. MONOMINERALIC AGGREGATES: QUARTZ After olivine, quartz is probably the silicate mineral whose crystal plastic behavior has been best characterized. There are extensive data on the processes and microstructures of dislocation creep in quartzites which have come from experiments in solid media (Griggs-type) apparatus, but fewer good mechanical and flow law data because quartzites have not been successfully deformed in a gas apparatus, due to the apparent requirement for water fugacities that can only be achieved at pressures >~1000 MPa. The first sections below briefly summarize results from experimental studies; they are followed by a section of comparisons with naturally deformed quartzites. Water A discussion of crystal plasticity of quartz must begin with a consideration of the role of water. The importance of trace amounts of water was first demonstrated for quartz by Griggs and Blacic (1964) and Griggs (1967) and termed ‘hydrolytic weakening’. The state of our knowledge about the form and effects of water on quartz properties and deformation has been reviewed by Paterson (1989) and Kronenberg (1994); a few major points and recent results are summarized here. At experimental conditions of 900°C, 1.5 GPa and 10-6/s, clear quartz crystals are extremely strong and experience brittle failure, whereas milky quartz crystals and natural quartz aggregates such as quartzites and novaculites are weak and undergo steady state dislocation creep. Infrared analyses show that natural milky quartz crystals as well as the grains of quartzites contain a high average concentration of water (~2000-10,000 ppm or H/106Si), largely in the form of freezable fluid inclusions associated with deformation microstructures such as healed cracks and dislocations (e.g., Kronenberg et al. 1990), whereas clear quartz crystals have a very low water content (~30 ppm). The OH concentration in hydrothermally grown synthetic quartz crystals, as well as in natural amethyst and citrine, can be large but most is in a ‘gel-form’ and is non-freezable. Although ‘wet’ synthetic crystals are weak (at least transiently) even at low pressure, it is not clear that they have the same properties or undergo the same deformation processes as the quartz in natural crustal aggregates. There is still debate about the water-related defect that causes weakening; Griggs (1967) postulated that hydrolyzed Si-O bonds reduce the Peierls stress for glide, whereas Hobbs (1984) postulated that water-related defects act as dopants which influence the concentration of charged defects. Intragranular water certainly facilitates the creep recovery processes; annealing experiments on pre-deformed quartzites show that the rates of both dislocation climb and grain boundary migration are enhanced by water and

56

Tullis

pressure (Tullis and Yund 1989). Climb depends on volume diffusion, and whereas oxygen volume diffusion in quartz is almost immeasurably slow in a dry CO2 environment (Sharp et al. 1991), in wet environments its rate increases linearly with water fugacity and the diffusing species appears to be molecular water (Farver and Yund 1991). Whatever the water-related defect may be, the climb-accommodated dislocation creep strength of quartzite depends linearly on water fugacity (Kohlstedt et al. 1995; Post et al. 1996), and failure to take this into account explains some of the previous difficulties in matching experimental flow law extrapolations to reasonable crustal strengths. The water content of natural quartzites (~0.1 to 0.2 wt %) is greatly in excess of equilibrium, and yet at experimental conditions where dislocation creep occurs, the creep strength can be lowered by adding ~0.2 wt % water to the sample environment (e.g., Kronenberg and Tullis 1984), although greater amounts cause fracturing and dissolution/reprecipitation (e.g., den Brok and Spiers 1991; Post and Tullis 1998). The added water does not result in measurably higher intragranular water contents (den Brok et al. 1994), suggesting that the water residing in fluid inclusions does not act as a free fluid phase to control the water fugacity (e.g., Cordier and Doukhan 1989). Removing the water originally present in quartzites by vacuum drying greatly strengthens them (e.g., Kronenberg and Tullis 1984), but the water can be reintroduced at a rate consistent with ‘wet’ oxygen diffusivity into non-equilibrium sites such as fluid inclusions, rapidly restoring the low strength of the original ‘as-is’ aggregate (Post and Tullis 1998). These results suggest that over the range of crustal conditions where quartz deforms by crystal plasticity, it will rapidly readjust to changing water fugacity conditions. Slip systems In most crustal environments quartz will be in the low (α) form, with trigonal symmetry, although at high temperatures it may be in the high (β) form, with hexagonal symmetry. In either case slip requires strong Si-O bonds to be broken, and the strength of dry crystals approaches the theoretical shear strength (e.g., Kronenberg et al. 1986). There are few data on the CRSS of different slip systems in quartz crystals which contain water with the same form and concentration as that in natural quartzites, although there have been several studies using synthetic crystals at very low pressures (Baeta and Ashbee 1969). A combination of these experimental data plus inferences from LPOs developed in experimentally and naturally deformed quartzites indicates that basal slip is the easiest system at lower temperatures, prism and rhomb become easier with increasing temperature, and prism [c] becomes operative only at very high temperatures. Recrystallization mechanisms and microstructures In quartz the recovery process for dislocation creep and the resulting mechanical behavior change with flow stress, as does the mechanism of dynamic recrystallization. Three different recrystallization mechanisms have been identified in experiments on quartzites (Hirth and Tullis 1992), resulting in different creep ‘regimes’ characterized by distinct microstructures (Fig. 3). The original documentation of the different regimes was based on samples deformed at constant strain rate to moderately low strains in axial compression, with extensive correlation of TEM and optical microstructures with the mechanical behavior (Fig. 4). The same material was used in MSC experiments to determine the flow law for the higher temperature regimes (Gleason and Tullis 1995) and to provide a preliminary test of published recrystallized grain size piezometers (Gleason and Tullis 1993). Recent shear experiments to much higher strains (γ ~ 7), also in each of the three regimes, have allowed detailed characterization of the LPOs including the effect of dynamic recrystallization (Heilbronner and Tullis 2002; Tullis and Heilbronner 2002). These experimental dislocation creep regimes are described below, in order of increasing temperature (decreasing flow stress).

Deformation of Granitic Rocks

Figure 3. (a) Schematic illustrations of the optical microstructures developed in quartz aggregates deformed in the three dislocation creep regimes, characterized by distinct mechanisms of dynamic recrystallization. Note that the size of the recrystallized grains increases from regime 1 to 2 to 3, as flow stress decreases with increasing temperature and/or decreasing strain rate. Subsequent porphyroclasts and finely recrystallized tails are developed in regime 1; homogeneously flattened original grains are only developed in regime 2; complete recrystallization occurs at low strain in regime 3. (b) Schematic illustration of grain boundary bulging recrystallization, driven by local differences in dislocation density across portions of original grain boundaries. (c) Schematic illustration of subgrain rotation recrystallization. Progressive glide and climb of dislocations into low-angle boundaries (lighter lines) results in increasing misorientation of subgrains, until at about 10° the boundaries have the properties of ‘high angle’ grain boundaries (darker lines).

Figure 4. Representative curves of differential stress vs. axial shortening strain for samples of Black Hills quartzite (original grain size ~100 μm) deformed at a confining pressure of 1.5 GPa in dislocation creep regime 1 (bulging recrystallization), regime 2 (sub-grain rotation recrystallization) and regime 3 (migration recrystallization). W871: 850°C, 1.5×10-5/s, as-is; W872: 900°C, 1.5×10-5/s, 0.17 wt % water added; W858: 900°C, 1.5×10-6/s, 0.17 wt % water added.

57

58

Tullis

Figure 5. Microstructures developed in quartz aggregates experimentally deformed in axial compression in dislocation creep regime 1. (a) Optical micrograph (crossed polars) of sample shortened 58% (parallel to short edge of photo) at 700°C, 10-6/s; note inhomogeneously deformed grains. The parallel cracks formed at low temperature during unloading. (b) TEM micrograph showing recrystallized grains from sample deformed at 850°C, 10-5/s, 45% strain; note variable dislocation densities and strain-induced grain boundary bulging.

Regime 1: Bulging recrystallization. In the lowest temperature portion of the dislocation creep regime, volume diffusion rates are slow and dislocation climb cannot keep pace with the imposed strain rate; extreme work hardening occurs, with the development of very high densities of straight dislocation segments as well as cells and tangles. Locally along grain boundaries the dislocation density contrasts are sufficient to drive grain boundary bulging (e.g., Bailey and Hirsch 1962) from the lower into the higher density areas. These bulges pinch off to form extremely small (~1-2 μm) recrystallized grains, which are initially dislocation-free and thus much weaker than the work-hardened original grains (Fig. 5b). In this regime (termed regime 1 by Hirth and Tullis 1992), deformation of quartzites with a grain size significantly larger than the final recrystallized grain size is characterized by an initial stress peak, resulting from work hardening of the original grains, followed by strain weakening, resulting from progressive recrystallization (Fig. 4). Mechanical steady state requires substantial recrystallization, and can only be achieved in low strain axial compression experiments by using aggregates such as novaculite with an initial grain size (d ~ 5 μm) closer to the final recrystallized grain size. A component of grain boundary sliding in the extremely fine recrystallized grains cannot be ruled out, although the recrystallized regions develop a strong LPO and TEM shows that some grains are dislocation-free while others have very high densities, consistent with a dynamic steady state. The optical microstructures characteristic of low to moderate strain samples include relatively equant porphyroclasts with deformation bands and strong undulatory extinction, and anastomosing zones of extremely fine (often not resolvable) recrystallized grains along their boundaries (Fig. 5a). Strain is strongly partitioned into the recrystallized grains once an interconnected grain boundary layer has formed, and further recrystallization progressively reduces the size of the porphyroclasts and results in further weakening. As the porphyroclasts experience decreasing stress and strain rate, they may undergo limited internal recovery. Sheared samples develop an S-C′ fabric, with lensoid porphyroclasts defining the foliation and zones of recrystallized grains defining extensional shear bands. Regime 2: Subgrain rotation recrystallization. At somewhat higher temperatures and/or slower strain rates (lower stress), dislocation climb becomes easy enough to accommodate the imposed strain rate; thus in this regime of dislocation creep (termed

Deformation of Granitic Rocks

59

regime 2 by Hirth and Tullis 1992) both the mechanical behavior and the recrystallization process are quite different. The free dislocation density remains lower due to climbassisted annihilation, and geometrically necessary dislocations are arranged in lowenergy walls or subgrain boundaries. Gliding dislocations do not become tangled, thus there is no work hardening and original grains are fairly homogeneously flattened with increasing strain. Dynamic recrystallization occurs dominantly by progressive misorientation of subgrain boundaries (‘subgrain rotation’ of Guillope and Poirier 1979) until they become effectively high angle grain boundaries at a misorientation of ~10° (Fig. 6b). Because the recrystallized grains have about the same density of free dislocations as the subgrains within the original grains, recrystallization in this regime does not result in weakening (Fig. 4). Steady state flow is achieved at low strains regardless of original grain size or degree of recrystallization.

Figure 6. Microstructures developed in quartz aggregates experimentally deformed in axial compression in dislocation creep regime 2. (a) Optical micrograph (crossed polars) of sample shortened 60% (parallel to short edge of photo) at 800°C, 10-6/s; note homogeneously flattened original grains, some with deformation lamellae. (b) TEM micrograph of recrystallized region from same sample; dislocation densities are uniform, and both subgrain and high-angle boundaries are present.

The optical microstructures characteristic of deformation in this regime include relatively homogeneously flattened original grains containing optically visible subgrains, and recrystallized grains with a size (~10 μm) approximately the same as that of the subgrains and significantly larger than that in regime 1 (Fig. 6a). Subgrain rotations tend to be greatest at original grain boundaries, where stresses and strains are somewhat larger, thus producing ‘core and mantle’ structures (e.g., White 1975) of flattened and polygonized original grains surrounded by a necklace of recrystallized grains. At somewhat lower temperatures original grains may become extremely elongated without being replaced by recrystallized grains, while at slightly higher temperatures a sample may become completely recrystallized. An SC fabric is produced in shear, with slightly inequant recrystallized grains oblique to the shear zone boundary (Dell’Angelo and Tullis 1989), defining a steady state foliation (Means 1981). Regime 3: Grain boundary migration recrystallization. At still higher temperatures or slower strain rates (lower stress), dislocation climb remains easy but grain boundary mobility is much higher, resulting in a different process of recrystallization and different microstructures. In this regime (regime 3 of Hirth and Tullis 1992) recrystallization occurs chiefly by grain boundary migration; although the driving force is low, due to small differences in dislocation density and subgrain boundary spacing, the grain boundary velocity is high (Fig. 7b). Mechanical (and microstructural) steady state is achieved at very low strain (Fig. 4).

60

Tullis

The optical microstructures characteristic of this regime are quite distinct from those produced in the lower temperature regimes. Complete recrystallization occurs at low strain, before significant flattening of porphyroclasts can occur, and the microstructure maintains a dynamic steady state with increasing strain (Fig. 7a). Subgrains rarely accumulate enough misorientation to produce high angle boundaries, but any new grains produced in this way increase their size by boundary migration. The average recrystallized grain size is significantly larger (~40-50 μm) than that in regime 2, and larger than the average co-existing subgrain size (as also observed in halite by Guillope and Poirier 1979).

Figure 7. Microstructures developed in quartz aggregates experimentally deformed in axial compression in dislocation creep regime 3. (a) Optical micrograph (crossed polars) of sample shortened 57% (parallel to short edge of photo) at 900°C, 10-6/s with 0.12 wt % water added; the sample is completely recrystallized. (b) TEM of sample deformed at same conditions but to 36% strain; note low dislocation density, and boundary of large recrystallized grain migrating toward subgrain boundary (arrow).

Transitions. It should be emphasized that the transitions between these dislocation creep regimes are not sudden but gradual (Fig. 8a). In regime 1, bulging recrystallization may occur along internal boundaries such as deformation bands or microfractures, in addition to original grain boundaries. In high strain regime 1 samples, after significant strain weakening has occurred, porphyroclasts begin to develop flattened shapes and subgrains due to the decreased stress and strain rate they experience. At low strains in regime 2 recrystallized grains form by grain boundary bulging (‘suturing’) because subgrains have not accumulated enough misorientation to form high angle boundaries. At higher strain in regime 2 at least an increment of grain boundary migration must occur, because recrystallized grains formed by subgrain rotation retain an equant shape and a steady state oblique foliation angle. Experiments so far have been limited to the lower temperature portion of regime 3, but there is no evidence for any difference in deformation behavior or microstructures in the α− vs. β−quartz fields. Trace amounts of water affect the operative slip systems as well as the ease of dislocation climb and grain boundary migration, and the regime transitions for wateradded quartzites are observed to occur at lower temperatures than those for ‘as-is’ samples (Hirth and Tullis 1992) (Fig. 8b). Experiments in a MSC and with controlled water fugacity are needed to better define the stresses of the regime transitions and how they depend on water. A transition from dislocation creep to grain size sensitive flow has been documented in synthetic aggregates deformed at 300 MPa and 1100-1200°C (Brodie and Rutter 2000). Dislocation creep was observed in aggregates with a grain size of 20 μm and 0.6 wt % H2O, but grain size sensitive creep could only be produced in very dry and

Deformation of Granitic Rocks

61

Figure 8. Plots of temperature vs. strain rate showing the location of the dislocation creep regimes for quartz aggregates deformed at a confining pressure of 1.2 GPa and (a) ‘as-is’ or (b) with 0.17 wt % water added. Note that the boundaries between the regimes are actually gradational. Circles represent regime 1, plus symbols regime 2, solid squares are transitional between regimes 2 and 3, and open squares represent regime 3. For as-is samples, the fields of semi-brittle faulting and semi-brittle flow are included. (Modified from Hirth and Tullis 1992, with additional data from Hirth and Tullis 1994.)

extremely fine-grained (0.4 μm) aggregates. Extrapolation of these data to natural conditions indicates that grain size sensitive creep of aggregates with d ~ 1 μm would require a stress of ~30 MPa at 500°C and thus is unlikely unless grain growth is prevented due to pinning by other phases. LPOs LPOs in experimentally deformed quartz aggregates have been measured using several techniques. Optical techniques such as the Universal (U) stage provide information only on the [c] axis orientations, but the computer integrated polarization (CIP) technique (Heilbronner and Pauli 1993, 1994) allows very small grain sizes to be measured and grain orientation maps to be prepared. X-ray techniques yield the complete volume-averaged LPO of all crystal axes for all grains present within an ~100 μm thick section; electron back-scatter diffraction (EBSD) also provides the complete crystal orientation, but on a grain-by-grain basis. Early X-ray determinations of the complete LPOs for quartzites deformed in axial compression over a range of conditions were generally consistent with U-stage measurements of [c] axes for the same samples. They showed a transition with increasing temperature from a broad maximum of [c] axes parallel to compression to small circle girdles of increasing opening angle, consistent with a shift from dominantly basal slip to increasing contributions from slip on other planes (Tullis et al. 1973). The X-ray measurements showed a big difference in concentration for the positive and negative rhombs, resulting from mechanical Dauphine twinning (Tullis and Tullis 1972). The high symmetry and relatively low strain of these early experiments, plus the inability to measure grain-by-grain orientations for small grain sizes, did not allow investigation of two major questions about LPO development, namely the effects of shear geometry and dynamic recrystallization. Dell’Angelo and Tullis (1989) made U-stage measurements of porphyroclast LPOs in low shear strain samples, and Gleason et al. (1993) attempted to determine the LPOs of recrystallized grains in axial compression samples for each of the three recrystallization mechanisms identified by Hirth and Tullis (1992), but neither study was fully satisfactory. Recently quartzite samples have been deformed to high strain in general shear in all three dislocation creep regimes (optical microstructures are shown in Fig. 9), and for all samples the [c] axis LPOs have been

62

Tullis

Figure 9. Optical micrographs of samples of Black Hills quartzite deformed in general shear at a confining pressure of 1.5 GPa in dislocation creep regimes 1 (bulging recrystallization), 2 (subgrain rotation recrystallization), and 3 (migration recrystallization), before annealing (left column) and after annealing at the deformation temperature for 4 days (right column). Shear plane is horizontal and shear sense is sinistral. Micrographs taken with circular polarization; light grains = c-axes parallel to plane of thin section and dark grains = c-axes parallel to viewing direction. W940: 850°C, 10-5/s, as-is, γ = 4.3; W946: 875°C, 10-5/s, 0.17 wt % water added, γ = 7.2; W920: 900°C, 10-5/s, 0.17 wt % water added, γ = 2.1; W935: 915°C, 10-5/s, 0.17 wt % water added, γ = 5.6. [Used by permission of the Geological Society of London. From Heilbronner and Tullis (2002), Fig. 12.]

Deformation of Granitic Rocks

63

measured separately for porphyroclasts and recrystallized grains (Heilbronner and Tullis 2002; Tullis and Heilbronner 2002) (Fig. 10). The resulting patterns are very similar to those observed in naturally deformed quartzites; a brief summary is given below. The [c] axis LPO developed in regime 1 shows only a slight change in pattern with progressive recrystallization. At low strain (γ ~ 2) the porphyroclasts develop a broad peripheral maximum rotated slightly against the sense of shear, with a small tail toward the center of the pole figure, while the recrystallized grains form two broad maxima ~20° to 30° from the periphery (Heilbronner and Tullis 2002) (Fig. 10). At higher strain (γ ~ 5) and complete recrystallization, there is a sharper and stronger peripheral maximum that extends inward about 30°, rotated slightly with the sense of shear (Tullis and Heilbronner unpubl.). Although axis orientations have not yet been measured, the approach of Schmid (1994, Fig. 7) suggests that this [c] axis pattern could result from a combination of basal plus rhomb slip. The [c] axis LPO developed in regime 2 also shows only a slight change in pattern with progressive recrystallization. At low strain (γ ~ 2) the LPO is dominated by the porphyroclasts, and is a broad peripheral maximum rotated against the sense of shear

Figure 10. Bulk and partial c-axis pole figures (LPOs) of the quartzite samples illustrated in Figure 9, sheared in dislocation creep regimes 1, 2 and 3 (left 3 columns) and after annealing at the deformation temperature for 4 days (right column). Measured area = 0.4 mm2 (~80 original grains) except for regime 3 high γ where a combination of 3 areas (~240 original grains) was used. Shear plane is horizontal and shear sense is sinistral; LPO maximum density is indicated to the lower right of each figure. [Used by permission of the Geological Society of London. From Heilbronner and Tullis (2002), Fig. 9.]

64

Tullis

(Dell’Angelo and Tullis 1989). The LPO for a sample completely recrystallized at high strain (γ ~ 7) is a strong peripheral maximum extending inward about 30°, rotated with the sense of shear (Heilbronner and Tullis 2002) (Fig. 10), again consistent with basal plus rhomb slip. Thus in regimes 1 and 2, the evolution of the LPO with increasing shear strain is similar, and there is only a minor difference in pattern for porphyroclasts and recrystallized grains. It should be noted that Gleason et al. (1993) reported a different role for recrystallization in regime 1 LPO development. In flint samples (containing ~1 wt % water) axially compressed at effectively regime 1-2 conditions, a transient maximum of [c] axes parallel to compression was produced by the growth to impingement of rhombbounded porphyroblasts. The porphyroblast growth was apparently facilitated by the special mobility of the rhomb planes in the presence of water, and it favored the ‘hard’ orientation with no resolved shear stress on base, prism or rhomb systems. After impingement, however, further strain caused deformation and recrystallization of the porphyroblasts, and development of a small-circle girdle of [c] axes very similar to that developed in axially compressed quartzites. Thus the selective growth of ‘hard’ grains may be an experimental oddity. The [c] axis LPO developed in quartzites sheared in regime 3 undergoes a profound change at γ ~ 3-4, associated with complete recrystallization (Heilbronner and Tullis 2002). At lower strains the LPO of porphyroclasts is a broad peripheral maximum rotated against the sense of shear, whereas the recrystallized grains define two maxima about 2030° from the periphery (Fig. 10). However at γ ~ 4 and above, with complete recrystallization, the LPO switches to a partial girdle with strong maxima at the intermediate strain axis Y and at 35-40° to Y (Heilbronner and Tullis 2002) (Fig. 10). Although axis orientations have not yet been measured, these [c] axis patterns suggest a switch from dominantly basal plus rhomb slip to dominantly prism and rhomb slip. Although TEM analyses such as those reported by Knipe and Law (1987) for a naturally sheared quartzite have not yet been done, optical orientation analyses suggest that grains originally oriented for basal slip accumulate more internal strain energy compared to grains originally oriented for prism and rhomb slip; the latter grains begin to recrystallize at γ ~ 4 and the recrystallized grains preferentially grow into and consume grains of other orientations, including those forming the peripheral maximum. For all three dislocation creep regimes, recent experimental studies show that although static annealing following deformation profoundly changes the grain size and shape, the LPO is little affected (Heilbronner and Tullis 2002) (Fig. 10). Flow laws and piezometers Flow law parameters for recrystallization-accommodated (regime 1) dislocation creep have not yet been determined; they will probably require higher stresses and thus confining pressures than are possible in gas apparatus. Because water is known to affect the recovery process of grain boundary migration (e.g., Tullis and Yund 1989), a dependence of the flow law on water fugacity is expected, but it may not have the same dependence as found for climb-accommodated creep in regimes 2-3. A preliminary calibration of the recrystallized grain size piezometer for bulging recrystallization (Bishop, unpubl. data; see Post and Tullis 1999) found a significantly different slope for grain size vs. stress than has been found for subgrain rotation and grain boundary migration recrystallization in other minerals (e.g., Guillope and Poirier 1979; van der Wal et al. 1993). Early flow laws for quartz (e.g., Parrish et al. 1976; Kronenberg and Tullis 1984;

Deformation of Granitic Rocks

65

Jaoul et al. 1984; Koch et al. 1989) cannot be trusted due to errors in the measured stresses. Recent flow laws for climb-accommodated dislocation creep (including regime 2 and part of regime 3) have been determined by Paterson and Luan (1990) and Luan and Paterson (1992) using synthetic aggregates in a gas apparatus at 300 MPa and by Gleason and Tullis (1995) using a fine-grained natural quartzite in a MSC in a Griggs apparatus at 1.5 GPa. Both groups found n = 4 although there was a considerable range in activation energy Q (from 137±34 to 223±56 kJ/mol); however, when adjusted for the difference in water fugacity these flow laws are in remarkably good agreement (Kohlstedt et al. 1995). Constraints from comparisons with natural quartzites suggest that an activation energy of 135±15 kJ/mol and a water fugacity exponent of one are most reasonable (Hirth et al. 2001). Also, Gleason and Tullis (1995) used ‘as-is’ quartzite samples in their flow law determination at 1.5 GPa, and later analysis indicates that the water fugacity was likely somewhat less than that given by the pressure (Hirth et al. 2001). Thus far there has not been a calibration of the recrystallized grain size piezometer for subgrain rotation or for grain boundary migration recrystallization in quartz, using molten salt assemblies. Previously determined recrystallized grain size piezometer relations (e.g., Mercier et al. 1977; Christie et al. 1980; Koch 1983) did not take account of different recrystallization mechanisms and used sample assemblies which introduced serious errors in the measured stresses (Gleason and Tullis 1993). In new MSC piezometer calibrations it would be desirable to test for the independent effects of temperature (e.g., de Bresser et al. 2001) and water fugacity. Comparisons with naturally deformed quartzites The dislocation creep microstructures described by Hirth and Tullis (1992) have been observed in naturally deformed quartz aggregates, in the same sequence with increasing temperature, although the absolute temperatures of the regime transitions differ for different strain rates (Fig. 11). In addition, quartzites naturally deformed at high

Figure 11. Plot of temperature vs. differential stress, illustrating the range of deformation conditions predicted for best fitting quartzite flow law parameters. Estimated stresses at the transitions between the dislocation creep regimes in the Ruby Gap duplex are shown by the shaded regions. Differential stresses were calculated using a stress exponent (n) of 4 and a water fugacity exponent (m) of 1. An activation energy (Q) of 135 kJ/mol enables the entire range of differential stresses to be achieved within the estimated deformation conditions of temperature and strain rate. [Used by permission of the editor of International Journal of Earth Sciences. From Hirth et al. (2001), Fig. 6b, p. 85.]

66

Tullis

temperatures show microstructures not yet achieved experimentally. For example, within the regime where grain boundary migration recrystallization is dominant, at lower temperatures migrating boundaries are pinned by impurity phases such as micas, as observed experimentally, whereas at higher temperatures they are able to engulf the second phase particles (e.g., Drury and Urai 1990). In addition, at upper amphibolite to granulite grade, depending on water availability, one observes ‘chessboard’ subgrains indicative of a component of prism [c] slip (e.g., Mainprice et al. 1986) and large, highly irregular ‘dissection’ grains (e.g., Jessel 1987; Stipp et al. 2002a). Kruhl (1996) has suggested that prism [c] slip indicates deformation in the beta quartz field and thus can be used as a geothermobarometer; however, quartzite shear experiments in the beta field show no evidence of chessboard subgrains or a [c] maximum sub-parallel to the shear direction (Heilbronner and Tullis 2002). In experimental samples the temperatures of the recrystallization mechanism transitions depend on the water content of the samples (Hirth and Tullis 1992), which is not surprising since dislocation creep flow strengths depend on water fugacity. There is some evidence that the temperature of the brittle-plastic transition, as well as those of the dislocation creep regime transitions, are higher in quartzose rocks naturally deformed at ‘dry’ conditions, where the fluid is CO2-rich (e.g., Passchier 1995; Selverstone 2001) or where extreme grain size reduction may have decreased the thickness of the grain boundary fluid phase (e.g., Pennacchioni and Cesare 1997). Microstructural recognition of the different recrystallization mechanisms allows one to infer the relative temperatures of a suite of samples; more quantitative constraints require additional experimental information such as calibrated piezometer relations. Of the potential microstructural piezometers of dislocation density, subgrain size and recrystallized grain size, the latter is probably the most useful, since dislocation density can only be evaluated using TEM and is easily changed by events after the deformation of interest, and the subgrain size varies greatly depending on where it is measured within porphyroclasts and how it is measured (optical vs. TEM). For quartz, the three distinctly different mechanisms of dynamic recrystallization probably have different piezometer relations, as already shown for two different mechanisms in halite (Guillope and Poirier 1979) and in calcite (Rutter 1995). As mentioned above, early published piezometers for quartz (e.g., Christie et al. 1980) did not take account of different recrystallization mechanisms and had significant stress errors; for samples deformed in a MSC in regime 2-3, Gleason and Tullis (1993) found that the Twiss (1977) relation gave a reasonable match to the experimental data, although there are questions about its theoretical basis (e.g., Poirier 1985; de Bresser et al. 2001). Possible complications in using quartz recrystallization microstructures as indicators of deformation conditions could result from post-deformation static annealing, involving grain growth, or deformation continuing as temperatures decrease, involving overprinting of the higher temperature microstructures. Early studies of quartzites naturally deformed at high temperatures sometimes attributed complete dynamic recrystallization in regime 3 to static annealing. However, post-deformation static annealing is only indicated if the recrystallized grains have equant shapes, straight boundary segments and 120° angles, and/or if the recrystallized grain size does not ‘match’ the porphyroclast microstructures, and/or if metamorphic minerals have grown across the foliation. Applications of recrystallized grain size piezometers have generally been deemed ‘safe’ for rocks up through greenschist grade, whereas many amphibolite grade rocks appear to show static coarsening (e.g., Dunlap et al. 1997; Zulauf 2001). Cases of lower temperature overprinting are usually easy to recognize, because a smaller recrystallized grain size (or small-scale suturing) is superposed on a larger one.

Deformation of Granitic Rocks

67

Early attempts to apply an experimental piezometer and flow law to help interpret the thermomechanical history of naturally deformed quartzites were made by Hacker et al. (1990, 1992). Their approach was sound, although the results are somewhat ques-tionable because of their dependence on flow laws and piezometer relations based on experiments in which the stress was probably greatly in error (e.g., Gleason and Tullis 1993). Several examples of recent studies of naturally deformed quartzites and quartz veins which have documented the different recrystallization microstructures and have used the more recent flow laws are described below. All of them, however, are limited by using the Twiss (1977) piezometer relation for all three re-crystallization mechanisms.

Figure 12. Optical microstructures of quartz mylonites from the Ruby Gap duplex (crosses polars). (a) Regime 1 microstructure showing patchy undulatory extinction and fine recrystallized grains along grain boundaries. (b) Regime 2 microstructure showing flattened original grains with optically visible subgrains and recrystallized grains of the same size. (c) Regime 3 microstructure with complete recrystallization and oblique foliation indicative of grain boundary migration recrystallization. [Used by permission of the editor of the Internation Journal of Earth Sciences. From Hirth et al. (2001), Fig. 1, p. 79.]

In the Heavitree quartzite of the Paleozoic mid-crustal Ruby Gap duplex of Australia, Dunlap et al. (1997) used the microstructurally identifiable regime transitions to restore a series of thrust slices. They observed microstructures ranging from bulging recrystallization (d < 20 μm) (together with pressure solution in micaceous layers), to subgrain rotation (d ~ 20-60 μm), and to grain boundary migration (d ~ 40-200 μm), all within a temperature interval of ~250 to 330°C, from sub-greenschist to very low greenschist facies (Fig. 12). Using the Twiss (1977) piezometer they inferred that the stress associated with the bulging to subgrain rotation transi-tion was ~100 MPa and that associated with the subgrain rotation to grain boundary migration transition was ~60 MPa. Thermochronologic constraints allowed them to estimate a strain rate of ~10-15/s, and using this rate in the flow law of Paterson and Luan (1990) gave good agreement with their inferred temperatures and stresses.

In a major strike-slip shear zone of Oligocene age in the Eastern Alps (Alto Adige, Italy), Stoeckert et al. (1999) used optical and TEM observations to document quartz microstructures indicating semi-brittle flow and pressure solution from ~280 to ~310°C, and dislocation creep from ~310 to 350°C. Using the piezometer of Twiss (1977) they inferred a stress of 160 MPa for a grain size of 7 μm at ~280°C, and a stress of 60 MPa

68

Tullis

for a grain size of 30 μm at ~330°C. Using these stresses and temperatures in the flow laws of Paterson and Luan (1990) gave a strain rate of 10-13/s to 10-14/s. The deformation conditions and the recrystallized grain sizes in the studies of Dunlap et al. (1997) and Stoeckert et al. (1999) are thus quite similar. In the Bohemian Massif, Zulauf (2001) examined quartz veins in metagraywackes deformed during Variscan shortening over a range of conditions from lowermost greenschist facies (~300°C) to amphibolite facies (>570°C). He observed bulging recrystallization (d 60o (i.e., ) melt will be isolated in pockets in grain boundaries, triple junctions and four-grain junctions in order to minimize the total amount of solid-melt interfacial area. This situation is illustrated in Figure 1. (ii) For 0 < 2 < 60o (i.e., ), melt is interconnected along triple junctions and through four1529-6466/0051-0005$05.00

122

Kohlstedt

grain junctions, as shown in Figure 2. Melt along grain boundaries will be trapped in isolated pockets. (iii) For 2 = 0o (i.e., ), melt wets not only the triple junctions and four-grain junctions but also the grain boundaries. It should be noted that Equation (1) applies strictly only to melt-mineral systems for which (ss and (sm are single-valued.

Figure 1. Melt distribution for a melt-solid system with 2 > 60 o. (a) 3-D view illustrates the presence of melt pockets trapped in grain boundaries, triple junctions and four-grain junctions. (b) Crosssectional view of melt pocket trapped in triple junction. Modified from Lee et al. (1991).

Figure 2. Melt distribution for a melt-solid system with 0 < 2 < 60 o. (a) 3-D view illustrates the presence of an interconnected melt network along triple junctions and through four-grain junctions poc kets with melt pockets trap ped in grain bound aries. (b) C ross-sectional view of melt-filled triple junction. Modified from Lee et al. (1991).

Partial M elting and Deformation

123

For the olivine + basalt system, the average dihedral angle is 2 . 35o. Careful examination of the scanning electron microscopy (SEM) images in Figure 3 demonstrates that every triple junction contains at least a small amount of melt, demonstrating the an interconnected network of melt exists along these triple junctions and through the four-grain junctions. A closer view of melt-filled triple junctions is presented in the transmission electron microscopy (TEM) images in Figures 4a and b.

Figure 3. Backscattered electro n images of partially mo lten sample of olivine + 7 vol % basalt. (a) and (b) Lower magnification images. (c) and (d) Higher magnification images of a region near the center of (a) and (b), respectively. In the higher magnification images, arrow mark grain b ound aries (gb), melt-filled triple junctions (m-f tj) and melt-filled grain boundaries (m-f gb). Mo dified from K ohlstedt (1992).

124

Kohlstedt

Figure 4. Dark-field transmission electron microcopy images illustrating the range of melt distribution in samp les of olivin e + basalt. (a) Melt fills the triple junctions formed by olivine grains. (b ) Tw o me lt-filled triple junctions are connected by a melt-filled grain boundary. No te the angular shape of the gra ins in contract with melt.

In real rocks, neither (ss nor (sm is single-valued. The grain boundary energy (ss increases from near zero for low-angle grain boundaries to ~1 J/m2 for high-angle grain boundaries in olivine (Cooper and Kohlstedt 1982). Furthermore, (sm is highly anisotropic for the olivine-basalt and most other mineral-fluid system; melt tends to wet low-index crystallographic faces but not high-index planes. Consequently, even though the average dihedral angle is o0o, basaltic melt wets a portion of the grain boundaries in dunite. Examples are highlighted in Figure 3c and d in the backscattered electron SEM images of samples of olivine + basalt. At higher magnification in the TEM image in Figure 4b, two melt-filled triple junctions are clearly connected by a melt-filled (melt-wetted) grain

Partial M elting and Deformation

125

boundary. The anisotropy in (sm is emphasized by the angular shape of many of the meltsolid interfaces. Melt does not penetrate low-angle grain boundaries, that is, the dihedral angle is relatively large. One example is shown in the TEM micrograph in Figure 5 in which strain contrast from dislocations along the low-angle boundary indicates the absence of melt.

Figure 5. Bright-field image of melt-filled triple junction formed at the intersection of a high-angle grain bound ary grains with a low-angle grain bound ary in an o livine + b asalt sample. Note the periodic strain contrast along the low-angle bo undary indicative of a wellorganized array of disloca tions in an olivine grain.

Debate exists over the possible presence of a very thin film (~1 nm thick) of melt along all of the grain boundaries in partially molten rocks. High-resolution imaging and chemical analyses obtained with an analytical TEM have led different researchers to conflicting conclusions. Drury and Fitz Gerald (1996) , Wirth (1996) and de Kloe et al. (2000) have interpreted high-resolution observations in terms of the presence of very thin melt films, while Kohlstedt and co-workers have reached the opposite conclusion also based on highresolution TEM images and chemical examination with high spatial resolution (Vaughan et al. 1984; Kohlstedt 1990; Hiraga et al. 2002). In Figure 6, the high-resolution lattice-fringe image of the interface between two grains in an aggregate of olivine + basalt indicates that the grains are in direct contact with one another without the presence of a thin melt film. Somewhat surprisingly, however, high-resolution chemical analyses of such boundaries reveal a signature quite different from that of olivine. The X-ray intensity profiles in Figure 7 expose the presence of small amounts Ca, Al and Ti at a grain boundary between two olivine grains even though high-resolution lattice fringe images indicate that no melt film is present (Hiraga et al. 2002). The chemical composition at the grain boundary, however, is quite different from that of basalt. Hence, based on the high-resolution TEM

126

Kohlstedt

Figure 6. (a) Bright-field T EM images of grain boundary in the partially molten lherzolite samp le containing 9 vol % melt. (b) Lattice fringe image of two grains meeting at an app arently melt-free grain bound ary. (c) Diffraction pattern from grain boundary region. Mo dified from H iraga et al. (2002).

Fig ure 7. X-ray intensity profiles across an olivine-olivine grain boundary in a partially molten lherzolite aggregate obtained by STEM/EDX analyses. (a) M ajor and m inor elements. (b) Ca, Al and Ti on an expanded scale. Modified from Hiraga et al. (2002).

Partial M elting and Deformation

127

images and chemical analyses, Hiraga et al. (2002) concluded that elements soluble in olivine only in very low concentrations have segregated to the grain boundaries, markedly altering the composition of the grain boundaries by not forming a thin amorphous layer. DEFORMATION OF PARTIALLY MOLTEN ROCKS Over the past twenty years, a number of researchers have investigated the hightemperature plastic (viscous) deformation of partially molten mantle rocks (Cooper and Kohlstedt 1984, 1986; Bussod and Christie 1991; Beeman and Kohlstedt 1993; Jin et al. 1994; Kohlstedt and Chopra 1994; Hirth and Kohlstedt 1995a,b; Kohlstedt and Zimmerman 1996; Zimmerman 1999; Mei and Kohlstedt 2000a,b). In a creep experiment on a partially molten rock, the strain rate is measured as a function of differential stress F, grain size d, melt fraction N, temperature T, and pressure P such that the flow law for steady-state deformation will have the generalized form (2) The dependence of on F and d is commonly written as a power law, while the dependence of on T and P is usually expressed as an exponential relationship common for thermally activated processes. The power-law equation describing the creep of a solid + melt system deformed under anhydrous (dry) conditions then has the form (3)

where Adry is a material-dependent parameter, ndry and pdry are the stress and grain size exponents, and Edry and Vdry are the activation energy and activation volume for creep. The function , which describes the dependence of strain rate on melt fraction, remains a subject of both experimental and theoretical exploration. A similar expression can be written to describe the creep behavior of a partially molten rock under hydrous (wet) conditions (4)

where is a function of water concentration COH. Recent experimental work on samples of olivine + basalt suggests that, at least to a first approximation, = = , reflecting the fact that the melt distribution is similar under anhydrous and hydrous conditions (Riley and Kohlstedt 1990). Experiments also demonstrate that with r . 1 (Mei and Kohlstedt 2000a,b; Karato and Jung 2002). Extrapolation of laboratory results to flow at depth in the Earth necessitates extension to significantly lower strain rates. Strain rates in laboratory experiments are typically on the order of 10-4 - 10-6 s-1, while strain rates associated with plastic deformation in the Earth are closer to 10-14 s-1. Consequently, it is necessary to carry out experiments at differential stresses larger than those appropriate for flow in the Earth (e.g., 10 MPa in the laboratory versus 0.1 MPa in the Earth) and/or temperatures that are higher than those characteristic of the Earth. If flow laws derived from laboratory experiments are to be extrapolated to geological conditions, the deformation mechanism is in the laboratory must be the same as

128

Kohlstedt

those in the Earth. Therefore, experiments are generally carried out over as wide a range of , F, d, N, T and P conditions as possible in order to study deformation in both the diffusional and the dislocation creep regime. If the results are expressed as viscosity, 0, (5) then 0 will independent of stress in the diffusional creep regime (i.e., Newtonian with ) but will increase with decreasing differential stress in the dislocation creep regime (i.e., non-Newtonian with with 3 < n < 5). Laboratory experiments on partially molten olivine + basalt aggregates (Mei et al. 2002) suggest that the effect of water on strain rate can be separated from the effect of melt. Hence, Equations (3) and (4) can be combined as follows: (6) Again, this equation reflects the observation that the wetting behavior of basalt in a peridotite matrix is the same or very nearly the same under hydrous and anhydrous conditions, at least up to pressures of 0.3 GPa. Empirically, at the relatively small melt fractions used in our experiments, the melt-fraction dependence can be described quite well by (Kelemen et al. 1997; Mei et al. 2002; Zimmerman and Kohlstedt, submitted) (7) where the value of " depends on both the creep mechanism and the melt-solid system under investigation. Creep results for fine-grained (~15 :m) lherzolite samples deformed under anhydrous conditions are plotted in Figure 8 (Zimmerman and Kohlstedt, submitted). These samples composed of 61.8% olivine, 26.0% enstatite, 10.0% chrome-diopside and 2.2% spinel were hot-pressed from a powdered mantle xenolith from Damaping, China. In the diffusional creep regime (Fig. 8a), " = 21, while in the dislocation creep regime (Fig. 8b), " = 30. Somewhat larger values for " have been reported for the olivine + basalt system (Kelemen et al. 1997; Mei et al. 2002). The exponential dependence of strain rate on melt fraction given by Equation (7) results in a greater sensitivity of creep rate to melt fraction than predicted theoretically. Cooper and Kohlstedt (1986) and Cooper et al. (1989) analyzed the contribution of melt to creep rate for the diffusion creep regime. They concluded that melt enhances creep rate in two distinct ways: (1) the presence of melt results in a local stress enhancement due to replacement of a fraction of the grain boundary contact, , by melt (Fig. 2); (2) melt in triple junctions effectively shortens the path through which ions must travel in diffusing from grain boundaries under maximum compressive stress to those under smaller compressive (or tensile) stress. This analysis is based on an ideal melt distribution dictated by Equation (1) with a single-valued dihedral angle with 0 < 2 < 60°, such that all of the melt is distributed in triple junctions and four-grain junctions but is absent along grain boundaries. The CooperKohlstedt model yields

Partial M elting and Deformation

129 (8)

where the subscripts sc and se indicate short-circuit transport and local stress enhancement, respectively, and (9)

For N / 0.05 Equation (8) underestimates the effect of melt fraction on strain rate (Hirth and Kohlstedt 1995a,b). This equation does not provide a good description of the creep of partially molten rocks in large part because of the difference between the observed melt

Figure 8. (a) Normalized strain rate versus melt fraction for samples of synthetic lherzolite deformed in the diffusiona l creep regime. C et al. = Coope r et al. (1989), Equation (8); H & K = Hirth and Kohlstedt (1995a), Equation (11); M et al. = M ei et al. (in press), Equation (7); Z & K = Zimm erma n and Ko hlstedt (su bmitted), Equation (7). (b) Normalized strain rate versus melt fraction fo r samp les of synthetic lherzo lite deformed in the dislocation creep regime. H & K = Hirth and Kohlstedt (1995b), Equation (14); T = Tharp (19 83), Equation (13); M et al., Equation (7); Z & K, Equation (7). Modified from Xu et al. (2002).

130

Kohlstedt

distribution and that the ideal microstructure assumed in the Cooper-Kohlstedt derivation (Cooper and Kohlstedt 1986; Cooper et al. 1989). The anisotropy in solid-melt interfacial energy results in a melt distribution in which melt is not simply confined to triple junctions and four-grain junctions but also penetrates a significant fraction of the grain boundaries and separates neighboring grains, as illustrated in Figures 3 and 4. As a result, the partially molten samples have short-circuit diffusion paths and stress-enhancement geometries not considered in the derivation of Equation (8). In their experimental investigation of diffusion creep of partially molten olivine + basalt aggregates, Hirth and Kohlstedt (1995a) accounted for this effect by quantifying the fraction of grain boundaries completely wetted by melt, $. Based on microstructural examination of samples with melt fraction between 0.02 and 0.12, they demonstrated that $ increases linearly with melt fraction: (10) To modify Equation (8) in order to account for the presence of wetted grain boundaries, Hirth and Kohlstedt (1995a) argued that the effective diffusion path must be shortened from to . Equation (8) thus becomes

(11)

It should be noted that, at small melt fractions, Equations (8) and (11) can be wellapproximated by exponential functions, thus potentially explaining why the dependence of the strain on melt fraction can be approximated by Equation (7) (Mei et al., in press). Hirth and Kohlstedt (1995a) determined a value of $ = 4.5 N for their aggregates of olivine + basalt deformed under anhydrous conditions, while Mei et al. (in press) determined a value of $ = 3 N for similar samples deformed under hydrous conditions. Zimmerman and Kohlstedt (submitted) have shown that Equation (11) fits the experimental data for partially molten lherzolite samples with a value of $ = 2 N, as illustrated in Figure 8a. The smaller value of $ reported for partially molten lherzolite is consistent with the observation that fewer graingrain contacts are wetted in the lherzolite samples than in the dunite samples. This differences arises because melt does wet some of the olivine-olivine grain boundaries but seldom wets pyroxene-pyroxene grain boundaries or olivine-pyroxene phase boundaries. Two models are generally considered in discussing the effect of a small amount of melt on the creep properties of partially molten rocks in the dislocation creep regime. Chen and Argon (1971) examined the stress enhancement produced by a weak second phase. Their analysis leads to the following expression for the dependence of strain rate on melt fraction: (12)

Tharp (1983) explored the effect of a small amount of porosity and obtained

Partial M elting and Deformation

131 (13)

with 0.98 # k # 2.26. Even for relatively low melt fractions, Equations (12) and (13) both significantly underestimate the influence of melt fraction on strain rate. Hirth and Kohlstedt (1995b) exploited the stress-enhancement portion of Equation (8) in order to account for the discrepancy between these models and the experimental observations for dislocation creep of olivine + basalt aggregates. Their analysis leads to (14) This modified form of Equation (8) applied to dislocation creep provides a reasonable fit to the experimental data in Figure 8b. However, it should be noted that Equation (14) does overestimate the strain rate at low melt fractions and underestimates the strain rate at higher melt fractions. Some part of the weakening due to the presence of melt is likely results from melt-assisted relaxation of stresses at melt-filled triple junctions that develop as a result of grain boundary sliding. MELT DISTRIBUTIONS IN DEFORMING ROCKS One of the exciting advances in the past few years in the study of deformation of partially molten rocks is the discovery of the pronounced redistribution of melt that occurs during deformation. This redistribution occurs on two scales. First, at the grain scale, melt in triple junctions develops a strong melt preferred orientation (MPO). As illustrated in Figure 9, this MPO is most pronounced in samples sheared to large strains (( / 1) (Kohlstedt and Zimmerman 1996; Zimmerman et al. 1999), although it is present in samples deformed in triaxial compression to smaller strains (g . 0.2) (Daines and Kohlstedt 1997). Second, at a larger scale, melt segregates to form melt-rich bands separated by melt-depleted regions. As illustrated in Figure 10, in laboratory experiments on samples with a grain size of ~10 :m, the melt-rich bands are ~50 :m wide and are separated by 100-200 :m (Holtzman et al., in press). Deformation-induced melt segregation does not occur in all partially molten rock samples. In the olivine + basalt system (Fig. 9), melt-rich bands do not form, even though a strong MPO develops during deformation. However, in samples of olivine + 25 vol% chromite + basalt, distinct melt-rich bands evolve rapidly (Holtzman et al., in press). Laboratory experiments demonstrate that melt segregation also takes place in the following systems: (i) olivine + Fe-S melt + basalt (Hustoft, private communication), (ii) olivine + albitic melt (Zimmerman, pers. comm.), and (iii) anorthite + melt (Ginsberg and Kohlstedt, in preparation). The transition from the grain-scale MPO to larger scale melt segregation appears to be associated with the relative values of the sample thickness, ‹th, and the compaction length, *c. The compaction length is the natural scale length that emerges in analyses of compaction during two-phase flow of a melt (fluid) phase through a deformable rock (matrix) (e.g., McKenzie 1984). Melt flow and matrix deformation are coupled over a distance of *c.

132

Kohlstedt

Figure 9. Reflected light o ptical m icrograph of sample of olivine + basalt deformed in simple shear to a strain of ( > 2. Note the strong MPO developed at ~25o to the she ar plane and antithetic to the shear direction. Also note tha t the melt remains distributed on the gra in scale. Modified from Zimmerman and K ohlstedt (1996).

Figure 10. Reflected light optical micrograph of sample of olivine + 25 vol% chromite + 6 vol% basalt deformed to a shear strain o f ( > 2. Note the melt-rich bands oriented at ~20o to the shear plane and antithetic to the shear direction.

Hence, melt migrates in response to deformation of the solid and eliminates pressure gradients in the solid. For *c on the order of the sample thickness, the melt distribution dynamically adjusts to the pressure gradients produced during shear deformation. The compaction length is a function of the sample permeability, k, the matrix viscosity, 0, and the melt viscosity, ::

Partial M elting and Deformation

133 (15)

If the lack of deformation-driven melt segregation in the olivine + basalt system is taken as a reference point, then the melt segregation in the other four systems listed above can interpreted in the following way. For the olivine + basalt system, *c / 10 mm, while ‹th . 1 mm, that is, *c n ‹th. (i) For the olivine + chromite + basalt and the olivine + Fe-S melt + basalt systems, chromite or Fe-S melt tends to reside in the basalt channels thus reducing the permeability, k, and consequently decreasing *c so that *c > ‹th. (ii) In the olivine + albitic melt system, the melt viscosity, :, is more than three orders of magnitude smaller than that in the olivine + basalt system. Hence, *c > ‹th. (iii) In the anorthite plus melt system, the matrix viscosity, 0, is more than an order of magnitude smaller than the matrix viscosity in the olivine + basalt system. Thus again, *c > ‹th. The phenomenon of melt segregation during pure shear deformation of partially a molten rock was explored theoretically by Stevenson (1989). In his analysis, the driving force for deformation-driven melt segregation is determined by the pressure gradients that develop due to spatial fluctuations in melt fractions. Regions that initially have smaller than the average amount of melt will be weaker than those regions that contain more melt. Consequently, for a rock deforming at a given strain rate, the mean pressure ( , where F1, F2 and F3 are the principal stresses) will be lower in the weaker regions than in the stronger regions. Melt therefore migrates to the weaker, melt-enriched regions further lowering their strength and enhancing the pressure gradients. A positive feedback situation is thus developed. Stevenson’s (1989) analysis did not yield scale lengths for the spacing and size of the resulting melt-rich bands, although he did conclude that the band spacing should be smaller than the compaction length. Further theoretical analyses and experimental studies are needed to fully characterize this important phenomenon. Nonetheless, the presence of deformation-driven melt segregation has several important implications for the physical properties of lower crustal and upper mantle rocks. First, once melt-rich bands develop, they provide zones of relatively low shear strength such as those observed in ophiolites (Kelemen and Dick 1995; Dijkstra and Drury 2002). Deformation will therefore localize on these bands, producing a distinct anisotropy to the viscosity structure. Second, because permeability increases as the square or cube of the melt fraction, melt-rich bands will be high permeability paths. These paths will permit rapid transport of melt and will result in a highly anisotropic permeability structure. Third, the presence of melt-rich bands at depth in the Earth should affect seismic properties such as polarization and shear wave splitting. In addition, melt-rich bands will influence seismic attenuation. Finally, similar remarks apply to other physical properties such as electrical conductivity. In summary, the presence of melt directly influences the physical properties of a partially molten rock. (i) The effect of melt on the viscosity of a partially molten rock is significantly larger than anticipated from current models because the melt distribution deviates substantially from that predicted for a simple solid + melt system. Specifically, because the solid-melt interfacial energy is highly anisotropic, a significant fraction of the grain-grain interfaces are wetted by the melt even though the dihedral angle is great than 0o. Hence, models based on an ideal melt distribution in which the melt is confined to triple junctions

134

Kohlstedt

and four-grain junctions break down, particularly at melt fractions great than ~0.05. (ii) A pronounced segregation of melt occurs during deformation of partially molten rocks. In laboratory experiments, segregation occurs in samples for which the compaction length is smaller than the sample thickness. The melt-rich bands that develop have a profound effect on the physical properties (rheological, transport, seismic) of the rocks. ACKNOWLEDGMENTS Support from the National Science Foundation through grants EAR-9906986, OCE0002463, EAR-0126277 and INT-0123224 is gratefully acknowledged. The author thanks Shun Karato for his careful reading of and insightful comments on this manuscript. REFERENCES Beeman ML , Kohlstedt DL (199 3) Deformation of fine-grained aggregates of olivine plus melt at high temperatures and pressures. J Geophys Res 98:6443-6452 Bussod GY Christie JM (1991) Textural development and melt top ology in spinel lherzolite expe rimentally deformed at hypersolidu s conditions. In J Petrol, Special Lherzolite Issue. Vol 32. M enzies MA, D upuy C, Nicolas A (eds) p 17-39 Chen IW, Argon AS (1979) Steady state power-law creep in hetero geneous alloys with microstructures A cta Metall 27:785-791 Coope r RF, Kohlstedt, D L (19 82) Interfacial energies in the olivine-b asalt system. In Advances in Earth and Planetary Sciences, Vol 12, High-Pressure Research in Geophysics. Akimoto S, Manghnani MH (eds) Center for Academic Publications, Tokyo, p 217-228 Cooper RF, Kohlstedt DL (1984) Solution-precipitation enhanced diffusional creep of partially molten olivine basalt aggregates du ring hot-pressing. Tecton ophys 107:207-233 Cooper RF, Kohlstedt DL (1 986) Rheology and structure o f olivine-basalt partial melts. J Geophys Res 91:9315-9323 Cooper RF, Kohlstedt DL, Chyung K (1 989 ) Solution-precipitation enhanced creep in solid-liquid aggregates which display a non-zero dihedral angle. Acta Metall 37:1759-1771 Daines MJ, Kohlsted t DL (19 97) Influence of deformation on melt topology in peridotites, J Geophys Res 102:10257-10271 de Kloe R, D rury M R, van Ro ermund H LM (2000) Evid ence for stable grain bo unda ry melt-films in experimentally deformed olivine-orthopyroxene rocks. Phys Chem Minerals 27:480-494 Dijkstra AH, Drury MR (in press) Microstructures and lattice fabrics in the Hilti mantle seciton (Oman Ophiolite): evidence for shear localization and melt weakening in the crust-mantle transition zone? J Geophys Res Drury MR , Fitz Gerald D F (199 6) Gra in bound ary melt-films in upp er ma ntle rocks. Geophys Res L ett 23:701-704 Hiraga T, A nderson IM , Zimm erma n M E, M ei S, K ohlsted t DL (in press) Struc ture and che mistry of grain boundaries in deformed olivine + basalt and partially molten lherzolite aggregates: evidence for melt-free grain boundaries. Contrib Mineral Petrol Hirth G, Kohlstedt DL (1995a ) Exp erimental constraints o n the dynamics of the p artially molten upper mantle: deformation in the diffusion creep regime. J Geophys Res 100:1981-2001 Hirth G, Kohlstedt DL (1995b) Experimental constraints on the dynamics of the partially molten upper mantle: 2. Deformation in the dislocation creep regime. J Geophys Res 100:15441-15449 Holtzman BK, Groebner NJ, Zimmerman ME, G insberg SB , Kohlstedt D L (in press) D eform ation-d riven m elt segregation in pa rtially molten rocks. Geo chem G eophys Geosys Jin ZM, Green HW II, Zhou Y (1994) The rheology of pyrolite across its solidus. EO S - Trans A m G eop hys Union 75:585 Karato SI, Jung H (in press) Effects of pressure on high-temperature dislocation creep of olivine. Phil Mag A Kelemen PB, Hirth G, Shimizu N, Spiegelman M, Dick HJB (1997 ) A review of melt migration processes in the adiabatically upwelling mantle beneath mid-oceanic spreading ridges. Phil Trans R Soc Lond A 355:283-318

Partial M elting and Deformation

135

Kelemen PB, Dick HJB (1995 ) Focused melt flow and localized deformation in the upper mantle: juxtaposition of replacive dunite and ductile shear zones in the Josephine peridotite SW O regon. J Geophys Res 100:423-438 Kohlstedt DL (1990) Chemical analysis of grain bound aries in an olivine-basalt aggregate using high-resolution, analytical electro n microsco py. In The Brittle-Ductile Transition in Rocks, the Heard Volume. Duba AG, Handin JW, W ang WF (eds) American Geophysical Union, Washington DC, p 211-218 Kohlstedt DL (1 992) Structure, rheology and permeab ility of partially molten rocks at low melt fractions. In Mantle Flow and Melt Generation at Mid-Ocean Ridges. Geophys Monogr 71, Phipps-Morgan J, Blackman DK, and Sinton JM (eds) American Geophysical Union, Washington DC, p 103-121 Kohlstedt DL, Chopra PN (1994) Influence of basaltic melt on the creep of polycrystalline olivine under hydro us conditions. In Magmatic Systems. Ryan MP (ed) Academic Press, New York, p 37-53 Kohlstedt DL, Zimmerman ME (1996) Rheology of partially molten mantle rocks. Ann Rev Earth Planet Sci 24:41-62 Lee VW , Mackwell SJ, Brantley SL (1991) Th e effect of fluid chemistry on wetting textures in novaculite. J Geophys Res 96:10023-10037 McKenzie DP (1984) The generation and compaction of partially molten rock. J Petrol 25:713-765 Mei S, Kohlsted t DL (2000a ) Influence of water on plastic deformatio n of olivine aggregates: 1. Diffusion creep regime. J Geophys Res 105:21457-21469 Mei S, Kohlstedt DL (2000b) Influence of water on plastic deformation of olivine aggregates: 2. Dislocation creep regime. J Geophys Res 105:21471-21481 Mei S, Bai W , Hiraga T , Kohlstedt DL (2002) Rheology of olivine-basalt aggrega tes under hydro us conditions. Earth Planet Sci Lett 201:491-507 Riley GN, Kohlstedt DL (1990) An ex perim ental stud y of melt migration in an olivine-m elt system. In Magma Transport and Storage. Ryan MP (ed) John Wiley & Sons, New York, p 77-86 Tharp TM (1983) Analogies between the high-temperature deformation of polyphase rocks and the mechanical behavior of porous powder metal. Tectonophys 96:T1-T11 Vaughan PJ, Ko hlstedt D L, W aff HS (1982) Distribution o f the glass p hase in hot-pressed, o livine-basalt aggregates: an electron microscopy study. Contrib Mineral Petrol 81:253-261 Smith CS (1948) Grains, phases and interfaces: an interpretation of microstructure. Trans AIME 197:15-51 Smith, CS (1964) Some elementary principles of polycrystalline microstructure. Metall Rev 9:1-47 Stevenson DJ (1989) Spontaneous small-scale melt segregation in partial melts undergoing deformation. Geophys Res Lett 16:1067-1970 von Bargen N , W aff HS (1986) Permeabilities, interfacial areas and c urvatures of partially molten systems: results of numerical computations of equilibrium microstructures J Geophys Res 91:9261-9276 Watson EB , Brenan JM, Baker D R (1990 ) Distrib ution o f fluids in the continental mantle. In Continental Mantle. Menzies MA (ed) Claredon Press, Oxford, p 111-125 W irth R (1996 ) Thin amo rpho us films (1-2 nm) at olivine grain bound aries in mantle xenoliths from San Carlos, Arizona. Contrib Mineral Petrol 124:44-54 Zimmerman ME (1999) The Structure and Rheology of Partially Molten Mantle Rocks. Ph.D. dissertation, University of M innesota, M inneapo lis

6

Dislocations and Slip Systems of Mantle Minerals Patrick Cordier Laboratoire de Structure et Propriétés de l'Etat Solide (ESA CNRS 8008) Université des Sciences et Technologies de Lille 59655 Villeneuve d'Ascq-Cedex, France INTRODUCTION

This chapter focuses on dislocations and slip systems that are responsible for the plastic flow of high-pressure mantle minerals. After briefly introducing some basic concepts on crystal plasticity, we describe some recent experimental advances of the last decade that have contributed to our understanding of the rheology of mantle minerals. Among them, we describe some progress in the achievement of deformation experiments under high pressure. A more detailed review is presented by Durham et al. (this volume). We present then a novel Transmission Electron Microscopy (TEM) technique: Large Angle Convergent Beam Electron Diffraction (LACBED), which is very well adapted to the characterization of dislocations in beam sensitive materials, as well as the determination of dislocation Burgers vectors and characters from the fine analysis of X-ray diffraction peak broadening. The present review is in the continuity of the paper entitled “Plastic Rheology of Crystals” by Poirier, published in 1995 in “Mineral Physics & Crystallography— A Handbook of Physical Constants” (AGU Reference Shelf 2). Poirier's paper reviewed the plasticity of some important minerals including olivine and pyroxenes. Concerning these two minerals, the reader is invited to refer to Poirier's contribution. In 1995, Poirier stated about high-pressure mantle minerals: “Despite their importance for the rheology of the transition zone and the lower mantle, there is no information on the plasticity of the high-pressure mantle minerals in the relevant conditions of temperature and pressure.” Seven years later, the situation has significantly evolved and is still rapidly changing. It is the aim of this paper to highlight these recent advances. BASIC CONSIDERATIONS Plastic flow is basically a transport phenomenon controlled by the motion of defects: point defects, dislocations, or grain boundaries. The driving force is provided by the applied shear stress. In many cases, deformation is produced by the motion of dislocations and the present paper will focus on this aspect only. The strain rate then obeys the Orowan's equation:

ε& = ρbv where ρ is the density of mobile dislocations, b is the length of the Burgers vector, and v is the dislocation velocity. A dislocation defines the boundary between the region of the crystal which has been sheared and the unslipped region in which the displacement is r zero (Fig. 1). A rdislocation is characterized by a line direction (unit vector u ) and its Burgers vector b which is the elementary amount of plastic shear carried by the dislocation. The Burgers vector is the most invariant characteristics of a dislocation, it is r r the same for all parts of the line. The angle between u and b defines the r dislocation r character: edge, screw or mixed (Fig. 2). The plane determined by u and b is called the glide plane (or slip plane), provided the dislocation line is not of pure screw character. A dislocation loop will sweep over its glide plane (Fig. 2) under the action of an applied stress (defined by the stress tensor σ ), the dislocation segments being subjected to the Peach-Koehler force: 1529-6466/00/0051-0006$05.00

138

Cordier

Figure 1. Slip of a crystal by propagation of an edge dislocation. The dislocation line visualized by an extra-half plane represents the border between the slipped area (on the left side) and the unslipped regions (on the right side). After Nicolas and Poirier (1976).

Figure 2. Expansion of a dislocation loop under the action of a shear stress applied on its glide plane. The edge and screw segments are indicated.

r r r F = b.σ × u

( )

In most cases, plastic deformation is due to the glide motion of the dislocations. At high temperature, point defects become mobile and can interact with dislocation to promote climb, i.e., motion of the dislocation line out of its glide plane. However, except in some special cases where plastic deformation proceeds by pure climb, strain is generally produced by dislocation glide (although the strain-rate might be controlled by thermally activated diffusion). The energy associated with a dislocation is mostly stored in the surrounding crystal as elastic energy. For a straight, mixed dislocation (character defined by θ, the angle between the line direction and the Burgers vector) contained in a cylinder of radius R, the energy per unit length is: E el =

μ b 2 (1− ν cos2 θ) 4π (1 − ν)

⎛R⎞ ln ⎜ ⎟ ⎝ ro ⎠

where μ is the shear modulus, ν is the Poisson ratio and r0 is the core radius. Macroscopic strain requires efficient dislocation multiplication mechanisms. A widely accepted mechanism for dislocation generation is based on the Frank-Read source (Fig. 3). In this model, a segment AB of length L bows out under an applied stress. The stress reaches a maximum σc when the loop is a semi-circle. If the stress exceeds σc the line develop into a complete loop restoring the initial segment AB (length L). The critical stress to activate a Frank-Read source is roughly: σc ∝

μb L

It is worth noticing that this critical stress is proportional to the shear modulus μ and that any change of this constant (for instance its evolution with applied pressure) will have a direct influence on the plastic properties.

Dislocations and Slip Systems of Mantle Minerals

139

Figure 3. The Frank-Read source. A segment AB of dislocation is represented in its glide plane. A and B represent fixed pinning points where the dislocation line leave the glide plane. Under an applied shear stress, the dislocation segments bows out and moves from position 1 to positions 2, 3 and 4 where the loop is about to close leaving in 5 a large glide loop and the original segment AB.

Figure 4. Schematic representation of an edge dislocation in a simple, monoatomic, cubic crystal. The Burgers vector and the glide plane are represented. The atoms of the extra-half plane are shown in white.

Figure 5. Movement of an edge dislocation. The dislocation originally in “1” moves by one interatomic distance to “3”. In its motion, the dislocation passes through an unfavorable configuration “2” which involves bond breaking.

While the elasticity theory describes correctly the shape of a dislocation at large distances from its line, some aspects are due particularly to the crystal structure. Among those is the structure of the dislocation core. Dislocations are usually represented on simple cubic lattices. Figure 4 represents the usual simplified representation of an edge dislocation in a simple cubic crystal. The extra-half plane is shown by the white atoms. Dislocation glide involves breaking bonds at the dislocation core without transport of matter (Fig. 5). In order to glide from position “1” to an identical position “3”, the line must go through an intermediary position “2” of higher energy. A stress σP must be applied to the dislocation to pass from “1” to “3”. This stress is a function of the glide plane considered and of the Burgers vector modulus. For a close-packed plane (large dhkl), the interatomic bonds are weak and not much distorted when one goes from “1” to “3”. The stress σP is going to be small. It is large, on the other hand, for not so close

140

Cordier

packed glide planes (small dhkl). A comprehensive treatment of this problem has been proposed by Peierls (1940) and Nabarro (1947), so σP is now called Peierls-Nabarro stress, or Peierls stress in short. Its expression is given by: σP ∝

2μ ⎛ 2π d hkl ⎞ ⎟ exp⎜ − ⎝ K b ⎠ K

where K depends on the dislocation character (K = 1 or 1-ν for screw and edge dislocations, respectively).

Left:

Figure 6. Schematic of an edge dislocation in a structure with two types of atomic planes. The dislocation core is made of two extra-half planes of both types. Center: Figure 7. The dislocation of Figure 6 is dissociated into two colinear partial dislocations separated by a stacking fault (glide dissociation). Right: Figure 8. Climb dissociation of the edge dislocation of Figure 6.

In fact, real dislocation cores are much more complicated. Complexity increases with the Burgers vector modulus (or with the number of atoms in the unit cell). When several atomic species are involved, edge dislocations exhibit, at least, two extra half-planes (Fig. 6). It might be energetically favorable that the dislocation dissociates in the glide plane into two partial dislocations separated by a stacking fault (Fig. 7). It is also possible that the dislocation dissociates by a climb process (Fig. 8). Climb dissociation occurs by conservative self-climb, which involves redistribution of matter between the partial halfplanes by short range diffusion. Mitchell et al. (1984) have shown that elasticity theory predicts the climb configuration to be usually more stable than the glide configuration, but climb dissociation requires significant diffusion. Very contrasted mobilities are expected to result from these two configurations. Glide dissociation is expected to enhance the mobility of the dislocations because their Burgers vector is smaller, and hence they are facing smaller Peierls stresses. Climb dissociation can have a marked effect on the mechanical properties of the mineral. In particular, an increase of the yield stress with increasing temperature can be expected (Fig. 9) in some cases. The curve σy(T) first decreases with increasing temperature. Figure 9. Possible evolution of the yield stress (σY) with temperature in a material in which dislocations are dissociated in climb. At low temperatures, the yield stress decreases with increasing temperature in a usual way. At intermediate temperatures the climb dissociation width increases resulting in a decrease of the mobility of dislocations. At high temperature, the climb dissociation width has reached its equilibrium value, but diffusion is active and dislocations can move by dragging the point defects associated to the fault.

Dislocations and Slip Systems of Mantle Minerals

141

The yield stress would begin to increase when climb dissociation appears. The dislocations must then recombine to be able to glide. This process is rendered more difficult when the dissociation width increases (i.e., when T increases). Above a given temperature, the fault can diffuse and the yield stress begin to decrease. This simple example illustrates the fact that taking into account the nature of the atomic bonds is not sufficient. The dislocation core fine structure can have a tremendous influence on the plastic properties. Some information about the dislocation core can be gained from Transmission Electron Microscopy (TEM) observation (especially using high resolution TEM or the high-resolution diffraction contrast mode: weak-beam dark-field), for instance on dissociation if the dissociation width is large enough. Below that scale, the core structure can only be calculated. More generally, common signatures of core effects are: unexpected deformation modes and slip geometries; strong and unusual dependence of flow stresses on crystal orientation and temperature; and, most commonly, a break-down of Schmid's law which states that glide on a given slip system commences when the resolved shear stress on that system reaches a critical value. One of the more fundamental question when considering plastic deformation of a given material is: what kind of slip system is activated, and why? This section is aimed to emphasize the main parameters which govern the choice of the slip systems. A few criteria are usually proposed to predict or account for the choice of the slip systems: 1. Glide planes are chosen among close-packed planes 2. Burgers vectors are chosen among the shortest lattice repeats 3. Easy slip systems correspond to minimum values of “b/dhkl” [Chalmers-Martius (1952) criterion] 4. The slip plane corresponds to the dissociation plane of the dislocations Let us examine these different conditions. Close-packed planes. This condition is quite straightforward in metals where all atoms have about the same sizes. In case of oxides or silicates, one notices that cations are much smaller than anions (oxygen). It seems thus reasonable to look for the possible glide planes based on the structure of the anionic sub-lattice. Indeed, some structures show a close-packing of oxygen. We will see that it is not a general rule, although it must be remembered that high-pressure structures tend to be composed of closest-packed arrays of atoms. In the case of ionic crystals, one can look for neutral planes, although it has been shown in oxides that this criterion is not always verified.

Silicates usually correspond to complex crystal chemistry involving several types of atomic bonding. The easiest glide planes are usually those in which shear breaks the weakest bonds. In silicates, at low pressure, it seems important to avoid shearing the strong SiO4 tetrahedra. A trend in mantle mineralogy is that increasing pressure increases coordination number resulting in the transformation of SiO4 groups into SiO6 groups. Prewitt and Downs (1998) note that an increase in coordination number is usually accompanied by a lengthening (as well as an increase of the ionic character) of the bonds. For example, R(SiIVO) = 1.62 Å in quartz while R(SiVIO) = 1.78 Å in stishovite. This might have important implication on the choice of the glide system as Prewitt and Downs point out that, as a rule of thumb, short bonds are the strongest, and long bonds are the weakest. If this trend was to be confirmed, the transformation of SiO4 groups into SiO6 group would not correspond to a strengthening of the bonds and the role of the structural units would be less marked in the choice of the glide planes.

142

Cordier

Short Burgers vectors. The Burgers vectors depend essentially on the Bravais lattice. The elastic energy of a dislocation is proportional to μb2 per unit length. It follows that dislocations with Burgers vectors chosen among the shortest lattice vectors will be strongly favored from the energetic point of view. It must be kept in mind however that when dislocations are formed by plastic deformation (i.e., by mechanisms similar to the Frank-Read source for instance), it may be the most mobile rather than the less energetic dislocations which multiply more rapidly (Nabarro 1967). Indeed, the activation of dislocations with the smallest Burgers vectors is not a law of nature and many examples can be found where easy slip involves dislocations with large Burgers vectors (For instance rutile, TiO2 which glides on {101} when , and would be shorter slip directions). In most cases, dissociation into partial dislocations is invoked to account for this behavior. Dissociated dislocations can have lower elastic energies per unit length; they might also be more mobile. However, Friedel (1964) points out that, in crystals with complex crystal chemistry, glide of partial dislocations might involve complicated atomic shuffles. Chalmers-Martius criterion. Chalmers and Martius (1952) have observed in several simple structures that easy slip systems correspond to minimum values of “b/dhkl”. Indeed, the atomic displacements at the dislocation core are then smaller, because they are spread along close-packed planes which exhibit larger dhkl. It is interesting to note that this criterion corresponds to the smallest Peierls stresses, that is, to dislocations which are expected to be more mobile in glide. The Chalmers-Martius criterion is more delicate to apply in minerals: which value must we consider for dhkl? Is it the actual interplanar spacing (dhkl sensu stricto) or the distance between occupied atomic layers? Once again, it can be anticipated that the validity of this criterion might be affected by dissociation and the nature of atomic bonding at the dislocation core. Dissociation plane. If dislocations are to dissociate in a given plane in a glide configuration, their mobility is expected to increase. A special attention must be given to stacking faults which do not modify the anionic sub-lattice, because they must correspond to relatively low surface energies. It has been shown in case of oxides that this criterion was difficult to assess. Moreover, oxides have revealed complicated schemes (Bretheau et al. 1979) like zonal dissociation (i.e., dissociation spread over several levels of atomic layers) or synchroshear (collective displacements of cations along directions which are different from the shear direction). EXPERIMENTAL ADVANCES

The recent progress accomplished in characterizing the deformation mechanisms of mantle minerals are largely due to considerable technical developments. In the first place, one must cite the developments of deformation experiments under high-pressure and high-temperature. This issue is reviewed by Durham et al. (this volume). We just intend in the present section to describe the procedure that we have used to induce plastic deformation of minerals at high-P, high-T using a multianvil apparatus. As far as mechanisms are concerned, TEM appears as a key technique to analyze the deformation microstructures. However, high-pressure minerals are often very sensitive to beam damage and detailed characterization of the dislocation Burgers vectors by the usual techniques of contrast extinction are often precluded by early amorphisation. We present here a technique for Burgers vector determination based on Large Angle Convergent Beam Electron Diffraction (LACBED), which is very well adapted to beam sensitive materials. The analysis of broadening of X-ray diffraction peaks is an alternative technique to get information on the dislocations at a mesoscopic scale. Some recent developments concerning the characterisation of dislocations are presented.

Dislocations and Slip Systems of Mantle Minerals

MgO

Crushable Alumina

Zirconia

Alumina

Pyrophyllite

Cr-doped MgO

LaCrO3 heater

Molybdenum

Thermocouple

Sample in Re capsule

143

Figure 10. (a, left) Schematic cross section showing details of the octahedral pressure cell used for deformation experiments in the multianvil apparatus. Hard alumina pistons placed on both ends of the specimen induce deviatoric stresses at high pressure. (b, right) Cross-section of a recovered 10 mm pressure cell after deformation at 22 GPa and room temperature. The ringwoodite sample is located in a rhenium capsule.

High-pressure deformation experiments

Gas- and solid-medium apparatuses that permit either constant strain rate or constant load deformation under controlled thermochemical conditions are only able to operate to pressures that correspond to the crust or uppermost upper mantle (0.7 and ~4 GPa, respectively). Investigation of deep mantle minerals requires higher pressures that can be generated by devices such as multianvil or diamond-anvil apparatuses. In diamond anvil cell experiments, plastic deformation is achieved by taking advantage of the pressure gradient within the cell (Sung et al. 1977; Meade and Jeanloz 1990) or by bridging the specimen between the two diamonds (Chai et al. 1998; Wenk et al. 2000). In multianvil experiments, plastic deformation is achieved by modifying the high-pressure assembly. We have used an experimental setup used previously up to 16 GPa (Fujimura et al. 1981; Green et al. 1990; Liebermann and Wang 1992; Bussod et al. 1993; Sharp et al. 1994; Cordier et al. 1996; Durham and Rubie 1997; Voegelé et al. 1998a) and have adapted it to smaller octahedron sizes (10/5, 10/4) in order to reach higher pressures (Cordier and Rubie 2001). Two hard alumina pistons are located at each end of the sample to make the assembly mechanically anisotropic (Fig. 10). The lower piston consists of solid alumina whereas the upper piston acts as the thermocouple sleeve and contains the thermocouple wires. Thin (25 µm) Re discs were inserted between the pistons and the polished faces of the sample to avoid chemical reaction between the sample and the pistons at high temperature. Cordier and Rubie (2001) emphasize the importance of using a two-step procedure in which synthesis of the high-pressure phases is performed independently from their deformation: 1. The first step corresponds to the synthesis of the high-pressure phase in the conventional quasi-hydrostatic cell assembly. After the synthesis experiment, the

144

Cordier

densified octahedron is broken to recover the sample capsule. The ends of the Re capsule are ground to expose the sample and it is characterized by Raman spectroscopy to ensure that the required phase has formed. 2. The second step corresponds to the deformation at high-P, high-T of the dense specimen of the high-pressure phase in the modified assembly (Fig. 10). The deformation experiment consists of increasing the pressure first (within three to four hours) at room temperature. During this cold compression stage, the size of the MgO octahedron decreases considerably (e.g., 20% shortening) because of the collapse of its porosity and flow between the WC cubes. The extruded material together with the pyrophyllite form the gaskets that contain the pressure medium. The pistons and sample are stiffer than the octahedron materials. This results in the development of high uniaxial stresses along the piston-sample column as the octahedron size decreases. After the required pressure is reached, temperature is increased to the desired value at a rate of 50-100°C/min. The uniaxial stresses then relax in the whole assembly, including in the sample, by plastic deformation. Following an annealing period (usually 30 min to 2 h, although much longer times are possible), the sample is quenched under pressure to dislocations interacting in their glide plane. Grain viewed down the c-axis. TEM weak-beam dark-field (g: 1010 ).

Figure 44. The unit cell of MgSiO3 perovskite described within the Pbnm space group.

Figure 45. (a) Ideal (cubic) perovskite structure viewed along [001] with a unit cell representation. (b) Structure of MgSiO3 perovskite viewed along the longest axis: [001]O (the subscript “o” stands for orthorhombic). An orthorhombic unit cell is represented to the right. The structure can be approximately described within a pseudo-cubic unit cell (shown on the left—labeled with subscript “pc”) with apc ≈ ao/ 2 ≈ bo/ 2 ≈ co/2.

Dislocations and Slip Systems of Mantle Minerals

163

Figure 46 (left). Dissociated (unrelaxed) [001](100) dislocation in the ideal, cubic, perovskite structure. The octahedra are figured with the oxygen ions represented as circles, the Mg cations are represented as squares (after Poirier et al. 1989). Figure 47 (right). Dissociated (unrelaxed) [ 11 0 ](110) dislocation in the ideal, cubic, perovskite structure. (after Poirier et al. 1989).

The possible dislocations in ideal cubic perovskites have been discussed by Poirier et al. (1983; 1989), who pointed out that the perovskite structure could be seen as a ReO3 structure stuffed with small cations at the center of the octahedra. Based on this analogy, they suggested that possible slip systems could be {001} and < 110 > c {1 10}c . Possible core structures including dissociation into two colinear partial vectors have been proposed (Figs. 46 and 47). Deformation experiments have been conducted on a large number of analogue materials: BaTiO3 (Doukhan and Doukhan 1986; Beauchesne and Poirier 1989), CaTiO3 (Doukhan and Doukhan 1986; Wright et al. 1992; Besson et al. 1996), SrTiO3 (Wang et al. 1993; Mao and Knowles 1996; Gumbsch et al. 2001), NaNbO3 (Wright et al. 1992), KTaO3 (Beauchesne and Poirier 1990) KNbO3 (Beauchesne and Poirier 1990), KZnF3 (Poirier et al. 1983), YAlO3 (Wang et al. 1999). It appears from these studies that perovskites do not constitute an isomechanical group. However, they confirmed that the dislocations in perovskite have c and c Burgers vectors. Moreover, the < 110 > c {1 10}c slip system seems to be activated from room temperature and controlled by glide, whereas the c{001}c slip systems would be characteristic of high temperature only and controlled by climb. A strength anomaly comparable to the one described in Figure 9 has been reported recently by Gumbsch et al. (2001) in SrTiO3 deformed in compression in the temperature range 100-1800 K. Microstructural investigations suggest that this behavior is related to climb dissociation of dislocations. Indeed, climb dissociation of dislocations had already been reported in SrTiO3 by Mao and Knowles (1996), and discussed by Poirier et al. (1983). The first deformation experiments on MgSiO3 perovskite have been performed at room temperature by Karato et al. (1990) using microindentation. They determined an average hardness Hv = 18±2 GPa. Coarse-grained polycrystals of MgSiO3 perovskite have been deformed at P = 24 GPa and T = 1000-1400°C in the multianvil apparatus using the procedure described in Cordier and Rubie (2001). Optical microscopy shows dense deformation bands which interact strongly with the twins of orthorhombic perovskite (Fig. 48). Given the high sensitivity of MgSiO3 perovskite to electronirradiation, the dislocation microstructure of these samples has been analyzed by T. Ungár and colleagues using the peak broadening method described above. Preliminary results show that both c and c dislocations are present in these samples with densities in the range 1.8-3.6 1012m-2. Magnesiowüstite (Mg,Fe)O. Magnesiowüstite (Mw) is likely to be the second most abundant phase of the lower mantle. It is the result of a large solid solution between periclase (MgO) and wüstite (FeO) which exhibits the same crystal structure. Although

164

Cordier

Figure 48. MgSiO3 perovskite deformed at 25 GPa, 1400°C in the multianvil apparatus. Optical micrograph showing orthorhombic twins sheared by dense deformation bands.

Figure 49. View of the unit cell of MgO.

its inferred volumetric ratio is not high enough to make Mw control the rheology of the lower mantle, one cannot exclude that there are regions of the lower mantle that are enriched in Mw. Mw is the only phase to keep the same crystal structure (“NaCl-type”) from lower mantle to laboratory conditions (Duffy et al. 1995). The Bravais lattice of Mw is fcc, space group Fm3 m . The lattice parameters of the end-members are: 0.4211 nm for MgO (Hazen 1976) and 0.43108 nm for FeO (Foster and Welch 1956). Mg+ (or Fe2+) and O2- are displaced by half a diagonal of the cube (Fig. 49). The predominant slip systems in NaCl-type crystals are 1/2{110} (Fig. 49). Each {110} plane contains a single slip direction, resulting in 6 possibilities. Among those, some slip planes cross at 90°, some others at 60 or 120°, resulting in different dislocation interactions. Slip is also possible over other close-packed planes such as {100} or {111}. When a dislocation moves on a {110} plane, ions of opposite sign approach each other at the core of the dislocation in a “midglide” position (Fig. 50). For dislocation movement on {100} planes on the other hand, Gilman (1959) has pointed out that strong repulsive forces develop because ions of like signs are forced to approach each other (Fig. 51). Using singlecrystal deformation experiments on periclase at around 0.63 Tm, Routbort (1979) showed that slip on 1/2{110} alone was about four times easier than combined slip on 1/2{110} and 1/2{001}. Indeed, for MgO, the Peierls regime is not observed down to 4 K on the {110} planes, while it is operating below 1200 K for the {001} planes (Fig. 52). Many experiments have been performed on single crystals loaded at room pressure along to activate the easy slip systems. Below 1300°C (0.52 Tm), optical birefringence shows that slip in any given region is restricted to a single 1/2{110} slip system; there is no extensive penetration between the various systems. Stress-strain

Dislocations and Slip Systems of Mantle Minerals

165

Figure 50. Slip on {110} in MgO. Left side shows two atomic layers of MgO in the unslipped position. Right side shows the same in the half-glided position along 12 [1 10].

Figure 51. Slip on {100} in MgO. Left side shows two atomic layers of MgO in the unslipped position. The right side shows the same in the half-glided position along 12 [011 ].

curves are characterised by a fairly high rate of strain-hardening. Between 1300 and 1700°C (0.52-0.65 Tm), slip on conjugate (90°) systems are able to interpenetrate whereas slip on systems making an angle of 60° are not (Day and Stokes 1964, Copley and Pask 1965). There is no dislocation reaction at the crossing of two conjugate slip systems because the dislocation reaction 1/2[ 1 10] + 1/2[110] → [100] is not favorable. Crossing just involves mutual cutting. From a geometrical point of view, the intersection between two conjugate {110} planes is parallel to the line of edge dislocations. The primary intersection to be considered is between screw dislocations in one plane and edges or screws in the other plane. The jog on a screw cannot move forward and leaves a cusp (eventually an edge dislocation dipole) in its wake.

166

Cordier

Figure 52. Critical resolved shear stress for MgO on {110} and {100}.

A jog on an edge dislocation can move conservatively with the dislocation although it is still required to move over the {100} plane. Above 1700°C (0.65 Tm), the specimens begin to harden again after 20-30% elongation in tension (Day and Stokes 1964). The shape evolution of the specimen shows that all slip systems (at 60 and 90°) can interpenetrate. This, together with the increased mobility of dislocation and point defects, permits the rearrangement and polygonization of dislocations. Dislocations belonging to slip systems at 60° can react favorably according to the reaction 1/2[110] + 1/2[0 1 1] → 1/2[101] to form a third dislocation lying parallel to . The resulting dislocation, which is pure edge, can move only in the {112} slip plane (Day and Stokes 1964; Copley and Pask 1965; Reppich and Hüther 1974). At low temperature, slip does not normally occur over this plane and the dislocation may be regarded as sessile. Typical dislocation structures after transient creep at ~1700°C (0.65 Tm) include dislocation walls and random network of dislocations in between (Hüther and Reppich 1973). With increasing deformation, the regions with random dislocation arrangements disappear and the parallel cell walls loose their preferred orientations and build up a nearly equiaxed cellular network. The cell boundaries act in two ways: firstly as “forest” for glide dislocations contributing to the work hardening, and as “sinks” (for dislocation annihilation) preventing the increase in flow stress. Thus, the rate-controlling processes occur in the dislocation network of the cell boundaries (Reppich and Hüther 1974). Loading along enhances the probability for multiple slip at high temperature because it results in a decrease of the resolved shear stress over the {110} planes relative to other close-packed planes. Taking τ011 as the resolved shear stress over the (011) plane, we have for the [001] tensile axis: τ100 = 0 and τ111 = 0.83 τ011 whereas for the [110] tensile axis:

Dislocations and Slip Systems of Mantle Minerals

167

τ100 = 1.41 τ011 and τ111 = 1.65 τ011 Below 1700°C (0.65Tm), the major slip systems (determined from the slip traces) are: {110} (Day and Stokes 1966; Clauer and Wilcox 1976) although evidence for slip on {001} and 1/2{111} is also reported (Clauer and Wilcox 1976). Slip systems at 90° could interpenetrate but those at 60° could not (Day and Stokes 1966). The wavy nature of the deformation bands suggests that considerable cross-slip took place during creep (Clauer and Wilcox 1976). Above 1700°C, slip on all systems could interpenetrate. Slip occurs over a variety of planes, but primarily over the plane of maximum macroscopic shear stress. Since this is an irrational plane, shear is accompanied by cross slip. Crystals pulled more than 60% above 1700°C completely recrystallized with a fine grain size. The grains grew on further annealing at 2000°C to yield simple polycrystalline specimens (Day and Stokes 1966). Deformation along results in an equal resolved shear stress on three of the six

1/2{001} slip systems and no stress on the 1/2{110} family. Yield stress for

this orientation is considerably greater than for the orientations and continuously decreased up to 1600°C (0.61 Tm) (Copley and Pask 1965). The ratio of the yield stresses is about 13:1 at 350°C (0.2 Tm), 4.5:1 at 600°C (0.29 Tm) and 3:1 at 1600°C (0.61 Tm). Creep experiments performed recently on Mg0.8Fe0.2O polycrystals at 0.44 Tm yielded CPO compatible with slip on both 1/2{110} and 1/2{001} although slip on {111} could not be excluded (Stretton et al. 2001).

Karki et al. (1997) have calculated the evolution of the elastic constants of MgO with pressure. They find dramatic change in elastic anisotropy with increasing pressure. We have calculated the anisotropic shear moduli μ[uvw](hkl) associated with shear along 1/2, parallel to {110}, {001} and {111} (Fig. 53). It is shown that 1/2{110} hardens considerably from the elastic point of view compared to 1/2{001}. 1/2{111} has an intermediate behavior. The influence of shear modulus on plasticity has been discussed by Poirier (1982). He remarks that the critical length for activating a Frank-Read source varies as μ/σ and that a decrease in modulus increases the dislocation density (more sources are available from the Frank network), causing an extra creep rate for the corresponding slip system. Oppositely, an increase of a shear modulus could decrease the activity of a slip system. Dramatic changes in slip systems can thus be anticipated for Mw at great depths with possible implications on the crystal preferred orientations and hence on the seismic properties. Garnets

Garnets represent an important mineral family for the rheological behavior of the upper mantle, and the mantle transition zone. They are often regarded as being very resistant to plastic flow. The structure of garnets is based on a bcc lattice (space group Ia3d). The unit cell (Fig. 54) contains eight structural units X3Y2Z3O12. It can be viewed as a threedimensional network of ZO4 tetrahedra (SiO4 in case of silicate garnets) and YO6 octahedra sharing corners and forming dodecahadral cavities filled with X cations. Therefore garnets comprise a family of complex silicates with widely varying chemical composition. Garnets with no Ca in the X site and Al in the Y site are called the pyralspite series. These consist of the end members: -

Pyrope -

Mg3Al2Si3O12

(a = 1.146 nm)

-

Almandine

Fe3Al2Si3O12

(a = 1.153 nm)

-

Spessartine

Mn3Al2Si3O12

(a = 1.162 nm)

168

Cordier

Figure 53. Evolution with increasing pressure of the shear moduli corresponding to the three slip systems 1/2{110}, 1/2{111} and 1/2{001} of MgO. The calculation is based on anisotropic elasticity, using the elastic constants proposed by Karki et al. (1997).

Figure 54. View of the unit cell of Mg3Al2Si3O12 pyrope.

Dislocations and Slip Systems of Mantle Minerals

169

Garnets with Ca in the A site are called the ugrandite series and consist of the end members: Ca3Cr2Si3O12 (a = 1.200 nm) - Uvarovite (a = 1.185 nm) - Grossularite Ca3Al2Si3O12 +3 (a = 1.206 nm) Andradite Ca3Fe 2Si3O12 In the transition zone, pyroxenes dissolve into the garnets resulting in a siliconenriched composition called majorite: Mg3(Mg, Si)2Si3O12 (= MgSiO3). The occurrence of Mg2+ cations (which are larger than the Al3+ cations) in the octahedral sites induces a distortion of the octahedron and in turn a displacement of the tetrahedra and of the dodecahedra. This results in a slight tetragonal distortion of the structure of the majorite which is described within the I41/a space group (Angel et al. 1989). The lattice parameters of MgSiO3 majorite are a = 1.1501 nm and c = 1.1480 nm with 32 formula units per cell. In cubic garnets, the shortest lattice repeat is 1/2 and, as in bcc metals, one can expect plastic deformation to be dominated by 1/2 dislocations. One must note however that, given the large unit cell of the garnet structure, such dislocation Burgers vector exhibit an unusually large magnitude (close to 1 nm). As in bcc metals, 1/2 dislocations from different slip systems are likely to react and form junctions following the reaction: 1 / 2[1 1 1]+ 1 / 2[1 1 1]→ [100]. Some earlier studies have given evidence that garnets could deform plastically in nature. Dalziel and Bailey (1968) and Ross (1973) used the distortions of Laue spots on X-ray diffraction photographs to infer for crystal plasticity in garnets from mylonitic rocks. Carstens (1969; 1971) has revealed the dislocations in pyrope garnets by etching the dislocation cores with a HF solution. The etch channels are then observed by optical microscopy. This method produced very clear pictures of the microstructures provided the dislocation density remains small enough. No information can be derived on the Burgers vectors using this method. Dislocation tangles and cell structures have been evidenced by this technique in pyropes from Norwegian and Czech peridotites (Carstens 1969). Several investigations on deformation microstructures in natural garnets have been carried out in the past few years using electron microscopy (Allen et al. 1987; Ando et al. 1993; Doukhan et al. 1994; Ji and Martignole 1994; Chen et al. 1996; Voegelé et al. 1998b, Prior et al. 2000). All these studies showed evidence for plastic deformation by dislocation creep. Characterizations of Burgers vectors lead to 1/2 and dislocations (Ando et al. 1993; Doukhan et al. 1994; Voegelé et al. 1998b). It is impossible however to decide in these samples whether slip has really been activated or if dislocations are junction products. Using LACBED, Voegelé et al. (1998b) have shown that 1/2 dislocations glide in three kinds of planes: {110}, {112} and {123}. The ease of glide seems to be close for these three families of planes although {110} seems to be the most frequently found. Dislocations often exhibit very marked crystallographic orientations: e.g., (70°, not screws ! see Fig. 55), and for 1/2 dislocations gliding in {110} planes (the easiest glide system). This marked anisotropy cannot be explained by elastic considerations, it must correspond to the structure of the core of the dislocations. LACBED studies on eclogites from Sesia Lanzo have revealed the occurrence of dislocations which have an uncommonly long Burgers vector (Voegelé et al. 1998b). These dislocations always appear to be dissociated (Fig. 56). Because they are stable at room pressure, most experimental studies on plastic deformation of garnets have been undertaken on non-silicate garnets: YAG (Hardiman et al. 1973; Parthasarathy et al. 1992; Corman 1993; Karato et al. 1994; Blumenthal and Phillips 1996), YIG (Rabier et al. 1976 a,b; Rabier 1979; Rabier et al. 1979; Rabier et al.

170

Cordier

Figure 55 (left). Garnet amphibolite from Bragance. Part of a 1/2[111] dislocation loop gliding in (1 1 0). Weak-beam dark field micrograph. g: 400 (courtesy V. Voegelé). Figure 56 (right). Eclogites from the Alps. dislocation dissociated in four partials (climb dissociation). Weak-beam dark field micrograph. g: 400 (courtesy V. Voegelé).

1981), GGG (Garem et al. 1982; Garem 1983; Garem et al. 1985; Wang et al. 1996) and CGGG (Voegelé et al. 1999). However, the plastic deformation of silicate garnets has been addressed using hardness tests (Karato et al. 1995a), creep (Karato et al. 1995a; Wang and Ji 1999) and high-pressure multianvil experiments (Voegelé et al. 1998a). Garnets exhibit a brittle-ductile transition around 1000°C at the timescale of laboratory experiments (Wang and Ji 1999; Voegelé et al. 1998a). Although this transition is shifted to lower temperature at natural strain-rates, it is suggested that ductile deformation of garnets is expected under mantle conditions only. Experimentally, 1/2 dislocations are found to glide in {110} (Wang and Ji 1999; Voegelé et al. 1998a), {112} and {123} (Voegelé et al. 1998a). dislocations compatible with the {011} and the {010} slip systems have also been identified (Voegelé et al. 1998a); see Figure 57.

Figure 57. Garnet deformed in the multianvil apparatus at 6.5 GPa, 900°C. [001] dislocation loops. Weak-beam dark field micrograph. g: 040 (courtesy of V. Voegelé).

Garnets represent an interesting case in crystal plasticity of a material deforming with dislocations having very large Burgers vectors (Nabarro 1984). This induces a very strong lattice-dislocation interaction (Peierls forces) as shown by the marked crystallographic orientations of the dislocation lines. Complex dislocation core structures, probably involving an extended, non-planar core, is probably responsible for the intrinsic lattice resistance to plastic shear. The core of such dislocations must expand and contract as the dislocations move from one Peierls valley to another (“breathing” of the core). Given the size expected for the core in such a structure as garnets, this renders dislocation

Dislocations and Slip Systems of Mantle Minerals

171

glide very unfavorable. The brittle, Peierls regime is found to operate up to high temperature: ~1000°C. Above this temperature the dislocations become mobile. This temperature is also that for the onset for recovery (as shown by the formation of subgrain boundaries) which shows that diffusion is efficient. Voegelé et al. (1998a) suggest that dislocation glide should be assisted by diffusion. Diffusion-assisted glide is still a conservative process. Local diffusion of point defects allows displacement of the sessile component of the dislocation core. The activation energy for the glide process is the migration energy of the point defects involved. 1/2 and slip operate in the whole temperature range. 1/2

and have also been characterized in majorite garnets of both synthetic and meteoritic origin (Voegelé et al. 2000). The case of majorite garnets provides another example for the influence of a small lattice distortion (here the tetragonal distortion from Ia3d to I41/a) on dislocations. In the present case, the Bravais mode is centered in both cases and both 1/2 and dislocations can be expected. However, the tetragonal distortion makes [100] and [001] dislocations having slightly different Burgers vector magnitudes and hence slightly different energies. However, these differences are not important enough to govern the dislocation activity. TEM observations do not suggest significant changes (such as dissociation) of the dislocation cores either. WHERE DO WE STAND?

A new field has emerged with the possibility of achieving deformation experiments under high-pressures exceeding ~10 GPa. Indeed, pressure appears clearly as the key parameter when dealing with the Earth's interior. The most important effect of pressure is probably the occurrence of the phase transformations that occur in the Earth's mantle. The importance of these phase transitions was recognized early on because they are critical to the understanding of the discontinuities observed by the seismologists. They also have a major effect on the rheology and hence on the dynamics of the mantle. We have seen that plastic deformation can occur through intracrystalline mechanisms (slip on crystallographic planes) that strongly depend on the crystal structure. This can have purely geometrical consequences such as the development of crystal preferred orientations that gives rise to seismic anisotropy (see Mainprice et al. 2000). The nature of dislocations and slip systems have direct consequences on the mechanical properties of the minerals and hence on the rheology of the mantle. Some structures appear very symmetrical and produce a large number of easy slip systems (garnets, MgO, ringwoodite or perovskite, for instance) when some other structures exhibit less degrees of freedom for slip (olivine, of course, which has only two orthogonal slip directions, but also wadsleyite and may be akimotoite). In some cases, the crystal structure is responsible for the activation of dislocations with very large Burgers vectors (wadsleyite, garnets). Such dislocations are clearly difficult to glide (lattice friction) and the mechanical properties are likely to depend strongly on whether glide dissociation is possible (wadsleyite) or not (garnets). Pressure also has a strong effect on the physical properties of the minerals, and among them, elasticity is one of the most important. Indeed, elasticity governs the long-range interactions between dislocations as well as the forces applied to them. We have seen in case of SiO2 or MgO that the evolution with pressure of the elastic properties is likely to have major effects on the rheology. The importance of the progress achieved should not lead us to forget the limitations of the studies so far conducted. In particular, it seems that high-pressure deformation experiments performed nowadays involve relatively high stress levels (compared to what is expected in the Earth's mantle). It is well known that dominant slip systems in a given material may change not only with temperature, but also with stress magnitude. This

172

Cordier

effect has been well demonstrated in case of olivine for instance ([001] vs. [100] glide). This problem raises the more general question of the extrapolation of laboratory data to natural conditions. This extrapolation involves scaling both in time and in space. There is no doubt that a detailed understanding of the physical processes involved in the plastic deformation of high-pressure minerals will be needed to address this important issue. The study on wadsleyite has pointed out the importance of dissociation on the choice of the slip systems. More generally, most high-pressure minerals exhibit a significant lattice friction that is probably governed by the dislocation core structure. It is important to realize that the lack of information about the core structure is probably the main drawback of our field at present. The dislocation core structure is crucial because it governs to a large degree the dislocation mobility. An example has been shown with SrTiO3 perovskite. In most cases the core structure is not accessible to direct characterization. It is certainly one of the major issues for the future to address the problem of dislocation core and mobility. We have discussed so far the influence on dislocation mobility of the lattice (friction) and of other dislocations (long- or short range interactions). The interaction with point defects or impurities can also have a strong influence on the mechanical properties. Among them is water, which is likely to be present everywhere in the mantle, albeit sometimes at the parts per million (ppm) concentration (Thompson 1992). The possible influence of water-related defects on plasticity has been shown in a spectacular way in the case of quartz (see Poirier 1995). The rheology of olivine is also affected by water although the effect is less marked in that case (Mackwell 1984; Mackwell et al. 1985; Mei and Kohlstedt 2000; Karato and Jung 2002). Beyond weakening, the effect of water might be to change the crystal preferred orientations (Jung and Karato 2001) with possible implications for seismic anisotropy. It is probably naive to infer from the quartz and olivine cases that water should weaken all mantle minerals in a similar way. It appears that water weakening involves very specific mechanisms which are already found to be very different in quartz and in olivine. It seems thus necessary to assess for each phase, the possible influence of water on the rheological properties. Some important issues have not been addressed in the present paper due to space limitations. One is related to diffusion. Although the importance of pressure has been highlighted, one must not forget that mantle conditions involve high temperatures that are likely to enhance atomic mobility. The first possible implication is the enhancement of dislocation climb, which might control the creep rate. In some case, plastic deformation can be achieved from the motion of point defects only. This mechanism is called diffusion creep. The absence of seismic anisotropy in the lower mantle has been interpreted as an evidence for deformation by diffusion creep (Karato et al. 1995b; McNamara et al. 2001). Modeling diffusion creep requires us to determine the diffusion coefficients of the relevant species at high-pressure and -temperature. The critical parameter for the relevance of diffusion creep versus dislocation creep is the grain size. Diffusion creep is likely to dominate for small grain sizes. A better knowledge of the grain size of mantle minerals is thus necessary to assess the possible role of diffusion creep in mantle deformation. The last issue we would like to address concerns the determination of mechanical data at high-pressure and -temperature. The major drawback of high-pressure deformation studies stems in the absence of mechanical data. Once again, this field has shown a rapid evolution in the past years. X-ray diffraction can be used to measure lattice strains from which elastic stresses can be deduced. This technique has been first applied to evaluate the strength of samples compressed in the diamond anvil cell (Kinsland and Bassett 1977; Meade and Jeanloz 1988a,b; 1990). Recently, it has been developed by the Stony Brook group in a multianvil apparatus connected to a

Dislocations and Slip Systems of Mantle Minerals

173

synchrotron beam line, see e.g., Weidner et al. (2001) and references therein and Durham et al. (this volume). For instance, Weidner et al. (2001) present preliminary mechanical data obtained on powdered samples for most phases that exist within a subducting slab which show that higher pressure phases tend to be stronger at elevated temperatures. These data provide clues for the evolution with depths of seismicity within a subduction zone. PERSPECTIVES: FROM ATOMIC TO THE GLOBAL SCALE

In their 1998 paper, Kubin et al. stated: “After being a major challenge in dislocation theory for decades, the multiscale modeling of plasticity is now becoming a reality.” The same kind of statement can be made concerning deformation experiments under extreme pressure and temperature, which makes it possible today to investigate the plastic behavior of minerals in P,T conditions of the deep mantle. Mechanical data obtained in situ from X-ray diffraction and microstructural investigation techniques performed postmortem provide information on the microscopic deformation mechan-isms. These new developments must be combined in a multidisciplinary approach to develop a forward modeling of the rheology of the mantle: o Ab initio or semi-phenomenological atomistic approaches can be applied to highpressure minerals to understand the core properties of dislocations. It is only at this scale that the effect of atomic bonding (and its evolution with pressure for instance) will be properly taken into account. Important constraints on the mobility of dislocations can also be expected from modeling at the atomic scale. On the other hand, simulated core properties will be constrained by TEM observations of dislocations in experimentally deformed samples. o The mesoscopic approach combines a simplified description of the core properties with the more rigorous elastic theory of dislocations with a view to understand the formation and dynamics of the microstructures of deformed specimens. Comparison with mechanical data obtained in situ will provide a check or a complement on the mobility laws calculated at the atomic scale. Mesoscopic simulations of dislocation dynamics and interactions are now available for both fcc and bcc metals (Kubin et al 1998). They are being developed for other crystal symmetries such as those encountered with deep mantle minerals. As an illustration, Figure 58 presents a preliminary simulation of [100](001) glide in olivine.

Figure 58. Simulated configuration corresponding to a crystal of forsterite deformed in single slip on [100](001) under a stress of 150 MPa applied along [101] and at 1300 K (Courtesy J. Durinck).

174

Cordier

o The continuum theory of solids can provide the adequate formulation of boundary value problems. For instance to go from the single crystal to the polycrystal (modeling of the finite strain field can provide the crystal preferred orientations needed to interpret the anisotropic seismic properties) or to go from single crystals to multiphase assemblages. This issue is reviewed by Dawson (this volume).

Plastic deformation of mantle minerals thus appears to be one of the most exciting and challenging issue of geophysics, which involves the leading aspects of solid state physics, materials science, high pressure mineral physics on both the theoretical and experimental sides. The goal is worth this effort. One can expect this multidisciplinary approach to provide a forward modeling of the rheology of the mantle which is needed to understand mantle convection at a global scale. ACKNOWLEDGMENTS

The author thanks S-i. Karato for an invitation to contribute to this volume and for many stimulating discussions. A large body of the work presented here has been performed at the Bayerisches Geoinstitut, Bayreuth during a sabbatical leave (“Congé thématique pour recherches”) from the University of Lille and subsequent short term visits. Continuous support from the European Union (“IHP—Access to Research Infrastructures” Programme, Contract No. HPRI-1999-CT-00004 to D.C. Rubie) and from CNRS-INSU (Programme “Intérieur Terre”) is gratefully acknowledged. The paper has benefited from careful reading and fruitful remarks from S-i. Karato and L. Kubin. REFERENCES Allen FM, Smith BK, Busek PR (1987) Direct observation of dissociated dislocations in garnet. Science 238:1695-1697 Ando J, Fujino K, Takeshita T (1993) Dislocation microstructures in naturally deformed silicate garnets. Phys Earth Planet Int 80:105-116 Andrault D, Fiquet G, Guyot F, Hanfland M (1998) Pressure-induced Landau-type transition in stishovite. Science 282(5389):720-724 Angel RJ, Finger LW, Hazen RM, Kanzaki M, Weidner DJ, Liebermann RC, Veblen DR (1989) Structure and twinning of single-crystal MgSiO3 garnet synthesized at 17 GPa and 1800°C. Am Mineral 74: 509-512 Beauchesne S, Poirier JP (1989) Creep of barium titanate perovskite: a contribution to a systematic approach to the viscosity of the mantle. Phys Earth Planet Int 55:187-199 Beauchesne S, Poirier JP (1990) In search of a systematics for the viscosity of perovskites: creep of potassium tantalate and niobate. Phys Earth Planet Int 61:182-198 Besson P, Poirier JP, Price GD (1996) Dislocations in CaTiO3 perovskite deformed at high temperature: A transmission electron microscopy study. Phys Chem Minerals 23:337-344 Blumenthal WR, Phillips DS (1996) High-temperature deformation of single-crystal yttrium-aluminum garnet (YAG). J Am Ceram Soc 79:1047-1052 Bretheau T, Castaing J, Rabier J, Veyssière P (1979) Mouvement des dislocations et plasticité à haute température des oxydes binaires et ternaires. Adv Phys 28:835-1014 Bussod GY, Katsura T, Rubie DC (1993) The large volume multi-anvil press as a high P-T deformation apparatus. Pure Appl Geophys 141:579-599 Carpenter MA, Hemley RJ, Mao HK (2000) High-pressure elasticity of stishovite and the P42/mnm–Pnnm phase transition. J Geophys Res 105B:10807-10816 Carstens H (1969) Dislocation structures in Pyropes from Norwegian and Czech garnet peridotites. Contrib Mineral Petrol 24:348-353 Carstens H (1971) Plastic stress relaxation around solid inclusions in pyrope. Contrib Mineral Petrol 32: 289-294 Chai M, Brown JM, Wang Y (1998) Yield strength, slip systems and deformation induced phase transition of San Carlos olivine up to the transition zone pressure and room temperature. In Properties of Earth and Planetary Materials at High Pressure and Temperature. Geophys Monogr 101:483-493 American Geophysical Union, Washington, DC Chalmers B, Martius UM (1952) Slip planes and the energy of dislocations. Proc R Soc London A 213:175-185

Dislocations and Slip Systems of Mantle Minerals

175

Chao ECT, Fahey JJ, Littler J, Milton DJ (1962) Stishovite, a new mineral from Meteor Crater, Arizona. J Geophys Res 67:419-421 Chao ECT, Littler J (1963) Additional evidence for the impact origin of the Ries basin, Bavaria, Germany. Geol Soc Am Abstr, p 127 Chen J, Wang Q, Zhai M, Ye K (1996) Plastic deformation of garnet in eclogite. Science in China (D) 39:18-25 Cherns D, Morniroli JP (1994) Analysis of partial and Stair-Rod dislocations by large angle convergent beam electron diffraction. Ultramicroscopy 53:167-180 Cherns D, Preston AR (1986) Convergent-beam diffraction of crystal defects. Proc 11th Intl Congr Electron Microscopy 1:721 Clauer AH, Wilcox BA (1976) High temperature tensile creep of magnesium oxide single crystals. J Am Ceram Soc 59:89-96 Copley SM, Pask JA (1965) Plastic deformation of MgO single crystals up to 1600°C. J Am Ceram Soc 48:139-146 Cordier P, Doukhan JC (1995) Plasticity and dissociation of dislocations in water-poor quartz. Philos Mag. A72:497-514 Cordier P, Morniroli JP, Cherns D (1995) Characterization of crystal defects in quartz by large-angle convergent-beam electron diffraction. Philos Mag A 72:1421-1430 Cordier P, Raterron P, Wang Y (1996) TEM investigation of dislocation microstructure of experimentally deformed silicate garnet. Phys Earth Planet Int 97:121-131 Cordier P, Rubie DC (2001) Plastic deformation of minerals under extreme pressure using a multi-anvil apparatus. Mater Sci Eng. A309:38-43 Cordier P, Sharp TG (1998) Large angle convergent beam electron diffraction determinations of dislocation Burgers vectors in synthetic stishovite. Phys Chem Minerals 25:548-555 Corman GS (1993) Creep of Yttrium Aluminium Garnet Single Crystals. J Mater Sci Lett 12:379-382 Dalziel IWD, Bailey SW (1968) Deformed garnets in a mylonitic rock from the greenville front and their tectonic significance. Am J Sci 266:542-562 Darot M, Gueguen Y (1981) High-temperature creep of forsterite single crystals. J Geophys Res 86(B): 6219-6234 Day RB, Stokes RJ (1964) Mechanical behavior of Magnesium Oxide at high temperatures. J Am Ceram Soc 47:493-503 Day RB, Stokes RJ (1966) Effect of crystal orientation on the mechanical behavior of Magnesium Oxide at high temperatures. J Am Ceram Soc 49:72-80 Doukhan N, Doukhan JC (1986) Dislocations in perovskites BaTiO3 and CaTiO3. Phys Chem Minerals 13: 403-410 Doukhan N, Sautter V, Doukhan JC (1994) Ultradeep, ultramafic mantle xenoliths—Transmission electron microscopy preliminary results. Phys Earth Planet Int 82:195-207 Duffy TS, Hemley RJ, Mao HK (1995) Equation of state and shear strength at multimegabar pressures: magnesium oxide to 227 GPa. Phys Rev Lett 74:1371-1374 Dupas C, Doukhan N, Doukhan JC, Green II HW, Young TE (1994) Analytical electron microscopy of a synthetic peridotite experimentally deformed in the β olivine stability field. J Geophys Res 99(B): 15821-15832 Dupas-Bruzek C, Sharp TG, Rubie DC, Durham WB (1998) Mechanisms of transformation and deformation in Mg1.8Fe0.2SiO4 olivine and wadsleyite under non-hydrostatic stress. Phys Earth Planet Interiors 108(1):33-48 Durham WB (1975) Plastic flow of single crystals of olivine. PhD dissertation, M.I.T., Cambridge, Massachusetts Durham WB, Froidevaux C, Jaoul O (1979) Transient and steady-state creep of pure forserite at low stress. Phys Earth Planet Int 19:263-273 Durham WB, Goetze C (1977) Plastic flow of oriented single crystals of olivine. 1. Mechanical Data. J Geophys Res 82:5737-5753 Durham WB, Goetze C, Blake B (1977) Plastic flow of oriented single crystals of olivine. 2. Observations and interpretations of the dislocation structures. J Geophys Res 82:5755-5770 Durham WB, Rubie DC (1997) Can the multianvil apparatus really be used for high-pressure deformation experiments? In Properties of Earth and Planetary Materials at High Presure and Temperature. MH Manghnani, Y Syono (eds) American Geophysical Union, Washington DC, p 63-70 Fei Y, Saxena SK, Navrotsky A (1990) Internally consistent thermodynamic data and equilibrium phase relations for compounds in the system MgO-SiO2 at high pressure and high temperature. J Geophys Res 95(B):6915-6928

176

Cordier

Foster PK, Welch AJE (1956) Metal oxide solutions: I. Lattice constants and phase relations in ferrous oxide (wustite) and in solid solutions of ferrous oxide and manganous oxide. Trans. Faraday 52: 1626-1634 Friedel J (1964) Dislocations. Pergamon Press, Oxford. Fujimura A, Endo S, Kato M, Kumazawa M (1981) Preferred orientation of β-Mn2GeO4. Programme and Abstract, The Seismological Society of Japan, p185 Garem H (1983) Dissociation des dislocations et plasticité de l'oxyde de structure grenat Gd3Ga5O12. Thèse de Doctorat d'Etat, University of Poitiers Garem H, Rabier J, Kirby SH (1985) Plasticity at crack tips in Gd3Ga5O12 garnet single crystals deformed at temperatures below 950°C. Philos Mag A 51:485-497 Garem H, Rabier J, Veyssière P (1982) Slip systems in gadolinium gallium garnet single crystals. J Mater Sci 17:878-884 Gillet P (1995) Mineral physics, mantle mineralogy and mantle dynamics. C R Acad Sci Paris 320:341-356 Gilman JJ (1959) Plastic anisotropy of LiF and other rocksalt-type crystals. Acta Metall Mater 5:608-613 Green II HW, Young TE, Walker D, Scholz CH (1990) Anticrack-associated faulting at very high pressure in natural olivine. Nature 348:720-722 Gueguen Y (1979) Les dislocations dans l'olivine des péridotites. Thèse de Doctorat d'Etat, Université de Nantes Gueguen Y, Darot M (1982) Les dislocations dans la forsterite déformée à haute température. Philos Mag A 45:419-442 Gumbsch P, Taeri-Baghbadrani S, Brunner D, Sigle W, Ruhle A (2001) Plasticity and an inverse brittle-toductile transition in strontium titanate. Phys Rev Lett 8708:5505-5508 Hardiman B, Bucksch R, Korczak P (1973) Observation of dislocations and inclusions in neodynium-doped yttrium aluminium garnet by transmission electron microscopy. Philos Mag 27:777-784 Hazen RM (1976) Effects of temperature and pressure on the cell dimension and X-ray temperature factors of periclase. Am Mineral 61:266-271 Horiuchi H, Hirano M, Ito E, Matsui Y (1982) MgSiO3 (Ilmenite-type): Single-crystal X-ray diffraction study. Am Mineral 67:788-793 Horiuchi H, Ito E, Weidner DJ (1987) Perovskite-type MgSiO3: Single-crystal X-ray diffraction study. Am Mineral 72:357-360 Horiuchi H, Sawamoto H (1981) β-Mg2SiO4: Single-crystal X-ray diffraction study. Am Mineral 66: 568-575 Hüther W, Reppich B (1973) Dislocation structure during creep of MgO single crystals. Philos Mag 28: 363-371 Jaoul O, Michaut M, Guegen Y, Ricoult B (1979) Decorated dislocations in forsterite. Phys Chem Minerals 5:15-19 ` Ji SC, Martignole J (1994) Ductility of garnet as an indicator of extremely high temperature deformation. J Struct. Geol 16:985-996 Jung HY, Karato S (2001) Water-induced fabric transitions in olivine. Science 293(:1460-1463 Karato SI (1990) Plasticity of MgSiO3 perovskite: the results of microhardness tests on single crystals. Geophys Res Lett 17:13-16 Karato S, Wang Z, Fujino K (1994) High-Temperature creep of Yttrium-Aluminium garnet single crystals. J Mater Sci 29:6458-6462 Karato S, Wang Z, Liu B, Fujino K (1995a) Plastic deformation of garnets: systematics and implications for the rheology of the mantle transition zone. Earth Planet Sci Lett 130:13-30 Karato SI, Zhang S, Wenk H-R (1995b) Superplasticity in Earth's lower mantle: evidence for seismic anisotropy and rock physics. Science 270:458-461 Karato S, Rubie DC (1997) Toward an experimental study of deep mantle rheology: A new multianvil sample assembly for deformation studies under high pressures and temperatures. J Geophys Res Solid Earth 102(B):20111-20122 Karato S, Dupas-Bruzek C, Rubie DC (1998) Plastic deformation of silicate spinel under the transitionzone conditions of the Earth's mantle. Nature 395:266-269 Karato SI, Jung H (2002) Effects of pressure on dislocation creep in olivine. Philos Mag A (in press) Karki BB, Stixrude L, Clark SJ, Warren MC, Ackland GJ, Crain J (1997) Structure and elasticity of MgO at high pressure. Am Mineral 82:51-60 Kingma KJ, Cohen RE, Hemley RJ, Mao HK (1995) Transformation of stishovite to a denser phase at lower-mantle pressures. Nature 374:243-245 Kinsland GL, Basset WA (1977) Strength of MgO and NaCl polycrystals to confining pressures of 250 kbar at 25°C. J Appl Phys 48:978-985 Kohlstedt DL, Goetze C (1974) Low-stress high-temperature creep in olivine single crystals. J Geophys Res 79:2045-2051

Dislocations and Slip Systems of Mantle Minerals

177

Kohlstedt DL, Goetze C, Durham WB (1976) New technique for decorating dislocations in olivine. Science 191:1045-1046 Kubin LP, Devincre B, Tang M (1998) Mesoscopic modelling and simulation of plasticity in fcc and bcc crystals: dislocation intersections and mobility. J Computed-Aided Mater Design 5:31-54 Langenhorst F, Poirier JP (2002) Transmission electron microscopy of coesite inclusions in the Dora Maira high-pressure metamorphic pyrope-quartzite. Earth Planet Sci Lett (submitted) Lee CY, Gonze X (1997) SiO2 stishovite under high pressure: Dielectric and dynamical properties and the ferroelastic phase transition. Phys Rev B Condensed Matter 56:7321-7330 Léger JM, Haines J, Schmidt M, Petitet JP, Pereira AS, Da Jornada JAH (1996) Discovery of hardest known oxide. Nature 383:401 Liebermann RC, Wang Y (1992) Characterization of sample environment in a uniaxial split-sphere apparatus. In High-Pressure Research: Application to Earth and Planetary Sciences. Y Syono, MH Manghnani (eds) American Geophysical Union, Washington, DC, p 19-31 Mackwell SJ (1984) Diffusion and weakening effects of water in quartz and olivine. PhD dissertation, Australian National University, Canberra Mackwell SJ, Kohlstedt DL, Paterson MS (1985) The role of water in the deformation of olivine single crystals. J Geophys Res 90(B):11319-11333 Madon M, Poirier JP (1980) Dislocations in spinel and garnet high-pressure polymorphs of olivine and pyroxene: implication for mantle rheology. Science 207:66-68 Madon M, Poirier JP (1983) Transmission electron microscope observation of α, β and γ (Mg,Fe)2SiO4 in shocked meteorites: planar defects and polymorphic transitions. Phys Earth Planet Int 33:31-44 Mainprice D, Barruol G, Ben Ismaïl W (2000) The seismic anisotropy of the Earth's mantle: from single crystal to polycrystal. In Earth's deep interior: mineral physics and tomography from the atomic scale to the global scale. Geophysical Monograph Vol. 117. SI Karato, AM Forte, RC Liebermann, G Masters, L Stixrude (eds) American Geophysical Union, Washington, DC, p 237-264 Mao Z, Knowles KM (1996) Dissociation of lattice dislocations in SrTiO3. Philos Mag A 73:699-708 Martini JEJ (1978) Coesite and Stishovite in the Vredeford dome, South Africa. Nature 272:715-717 McNamara AK, Karato SI, vanKeken PE (2001) Localization of dislocation creep in the lower mantle: implications for the origin of seismic anisotropy. Earth Planet Sci Lett 191:85-99 Meade C, Jeanloz R (1988a) The yield strength of MgO to 40 GPa. J Geophys Res 93:3261-3269 Meade C, Jeanloz R (1988b) The yield strength of B1 and B2 phases of NaCl. J Geophys Res 93: 3270-3274 Meade C, Jeanloz R (1990) The strength of mantle silicates at high pressures and room temperature: implications for the viscosity of the mantle. Nature 348:533-535 Mei S, Kohlstedt DL (2000) Influence of water on plastic deformation of olivine agregates 1. Diffusion creep regime. J Geophys Res 105:21457-21469 Mitchell TE, Lagerlöf KDP, Heuer AH (1984) Dislocations in ceramics. In Proceedings of the conference to celebrate the fiftieth anniversary of the concept of dislocations in crystals. The Institute of Metals, London, p 349-358 Mitchell TE (1999) Dislocations and mechanical properties of MgO-Al2O3 spinel single crystals. J Am Ceram Soc 82:3305-3316 Morniroli JP (2002) Large-Angle Convergent-Beam Electron Diffraction (LACBED)—Application to crystal defects. Société Française des Microscopies, Paris Morniroli JP, Cordier P, VanCappellen E, Zuo JM, Spence J (1997) Application of the Convergent Beam Imaging (CBIM) technique to the analysis of crystal defects. Microsc Microanal Microstruct 8: 187-202 Nabarro FRN (1947) Dislocations in a simple cubic lattice. Proc Phys Soc London 59:256-272 Nabarro FRN (1967) Theory of crystal dislocations. Oxford University Press, Oxford, England. Nabarro FRN (1984) Dislocation cores in crystals with large unit cells. In Dislocations 1984. P Veyssière, LP Kubin, J Castaing (eds) CNRS, Paris Nicolas A, Poirier JP (1976) Crystalline plasticity and solid state flow in metamorphic rocks. John Wiley & Sons, Ltd, London. Parthasarathy TA, Mah T, Keller K (1992) Creep mechanism of polycrystalline Yttrium Aluminum Garnet. J Am Ceram Soc 75:1756-1759 Peierls RE (1940) On the size of a dislocation. Proc Phys Soc London 52:34-37 Poirier JP (1982) On transformation plasticity. J Geophys Res 87:6791-6797 Poirier JP, Beauchesne S, Guyot F (1989) Deformation mechanisms of crystals with perovskite structure. In Perovskite: A structure of great interest to geophysics and materials science. A Navrotsky, D Weidner (eds) American Geophysical Union, Washington DC, p 119-123 Poirier JP, Peyronneau J, Gesland JY, Brebec G (1983) Viscosity and conductivity of the lower mantle: An experimental study on a MgSiO3 perovskite analogue, KZnF3. Phys Earth Planet Int 32:273-287

178

Cordier

Poirier JP (1995) Plastic rheology of crystals. In Minerals Physics and Crystallography—A Handbook of Physical Constants. AGU Reference Shelf 2. American Geophysical Union, Washington, DC, p 237-247 Presnall DC (1995) Phase diagrams of Earth-forming minerals. In Minerals Physics and Crystallography— A Handbook of Physical Constants. AGU Reference shelf 2. American Geophysical Union, Washington, DC, p 248-268 Prewitt CT, Downs RT (1998) High-pressure crystal chemistry. Rev Mineral 37:283-317 Prior DJ, Wheeler J, Brenker FE, Harte B, Matthews M (2000) Crystal plasticity of natural garnet: New microstructural evidence. Geology 28:1003-1006 Rabier J (1979) Dissociation des dislocations dans les oxydes de structure grenat. Application à l'étude de la déformation plastique du grenat de fer et d'yttrium (YIG). Thèse de Doctorat d'Etat, University of Poitiers Rabier J, Garem H, Veyssière P (1976a) Transmission electron microscopy determinations of dislocation Burgers vectors in plastically deformed yttrium iron garnet single crystals. J Appl Phys 47:4755-4758 Rabier J, Veyssière P, Grilhé J (1976 b) Possibility of stacking faults and dissociation of dislocations in the garnet structure. Phys Stat Sol (a) 35:259-268 Rabier J, Veyssière P, Garem H (1981) Dissociation of dislocation with a/2 Burgers vectors in YIG single crystals deformed at high temperature. Philos Mag A 44:1363-1373 Rabier J, Veyssière P, Garem H, Grilhé J (1979) Sub-grain boundaries and dissociations of dislocations in yttrium iron garnet deformed at high temperatures. Philos Mag A 39:693-708 Renner J, Stockhert B, Zerbian A, Roller K, Rummel F (2001) An experimental study into the rheology of synthetic polycrystalline coesite aggregates. J Geophys Res Solid Earth 106(B):19411-19429 Reppich B, Hüther W (1974) Formation and structure of dislocation networks developed during high temperature-deformation of MgO. Philos Mag30:1009-1021 Ringwood AE, Major A (1970) The system Mg2SiO4-Fe2SiO4 at high pressures and temperatures. Phys Earth Planet Interiors 3:89-108 Ross JV (1973) Mylonitic rocks and flattened garnets in the southern Okanagan of British Columbia. Can J Earth Sci 10:1-17 Ross N, Shu J-F, Hazen RM (1990) High pressure crystal chemistry of stishovite. Am Mineral 75:739-747 Routbort JL (1979) Work Hardening and creep of MgO. Acta Metall 27:649-661 Sharp TG, Bussod GY, Katsura T (1994) Microstructures in β-Mg1.8Fe0.2SiO4 experimentally deformed at transition-zone conditions. Phys Earth Planet Int 86:69-83 Sharp TG, Lingemann CM, Dupas C, Stöffler D (1997) Natural occurrence of MgSiO3-ilmenite and evidence for MgSiO3-perovskite in a shocked L chondrite. Science 277:352-355 Shieh SR, Duffy TS, Li B (2001) Strength of SiO2 across the stishovite-CaCl2 phase boundary. International Conference on High Pressure Science and Technology (AIRAPT-18) Shoemaker EM, Chao ECT (1961) New evidence for the impact origin of the Ries basin, Bavaria, Germany. J Geophys Res 66:3371 Sinclair W, Ringwood AE (1978) Single crystal analysis of the structure of stishovite. Nature 272:714-715 Stishov SM, Popova SV (1961) A new dense modification of silica. Geochem. (USSR) 10:923-926 Stretton I, Heidelbach F, Mackwell S, Langenhorst F (2001) Dislocation creep of magnesiowüstite (Mg0.8Fe0.2O). Earth Planet Sci Lett 194:229-240 Sung C-M, Goetze C, Mao H-K (1977) Pressure distribution in the diamond anvil press and shear strength of fayalite. Rev Sci Instrum 48:1386-1391 Tanaka M, Ueno K, Harada Y (1980) Large-angle convergent beam electron-diffraction. J Electron Microsc 29:277-277 Thompson AB (1992) Water in the Earth's upper mantle. Nature 358:295-302 Thurel E (2001) Etude par microscopie électronique en transmission des mécanismes de déformation de la wadsleyite et de la ringwoodite. PhD dissertation, University of Lille, Lille Tomioka N, Fujino K (1997) Natural (Mg,Fe)SiO3-ilmenite and -perovskite in the Tenham meteorite. Science 277:1084-1086 Tomioka N, Fujino K (1999) Akimotoite, (Mg,Fe)SiO3, a new silicate mineral of the ilmenite group in the Tenham chondrite. Am Mineral 84:267-271 Tsuchida Y, Yagi T (1989) A new, post-stishovite high-pressure polymorph of silica. Nature 340:217-220 Ungár T (2001) Dislocation densities, arrangements and character from X-ray diffraction experiments. Mater Sci Eng A Struct Mater 309:14-22 Ungár T, Borbély A (1996) The effect of dislocation contrast on X-ray line broadening: a new approach to line profile analysis. Appl Phys Lett 69:3173-3175 Ungár T, Dragomir I, Révész A, Borbély A (1999) The contrast factors of dislocations in cubic crystals: the dislocation model of strain anisotropy in practise. J Appl Crystallogr 32:992-1002

Dislocations and Slip Systems of Mantle Minerals

179

Ungár T, Tichy G (1999) The effect of dislocation contrast on X-ray line profiles in untextured polycrystals. Phys Stat Sol (a) 171:425-434 Voegelé V, Ando JI, Cordier P, Liebermann RC (1998a) Plastic deformation of silicate garnets I. Highpressure experiments. Phys Earth Planet Interiors 108:305-318 Voegelé V, Cordier P, Sautter V, Sharp TG, Lardeaux JM, Marques FO (1998b) Plastic deformation of silicate garnets II. Deformation microstructures in natural samples. Phys Earth Planet Interiors 108:319-338 Voegelé V, Liu B, Cordier P, Wang Z, Takei H, Pan P, Karato SI (1999) High temperature creep in a 2-3-4 garnet: Ca3Ge2Ge3O12. J Mater Sci 34:4783-4791 Voegelé V, Cordier P, Langenhorst F, Heinemann S (2000) Dislocations in meteoritic and synthetic majorite garnets. Eur J Mineral. 12:695-702 Wang Z, Karato S, Fujino K (1996) High temperature creep of single crystal gadolinium gallium garnet. Phys Chem Minerals 23:73-80 Wang ZC, DupasBruzek C, Karato S (1999) High temperature creep of an orthorhombic perovskite— YAlO3. Phys Earth Planet Interiors 110:51-69 Wang ZC, Ji SC (1999) Deformation of silicate garnets: Brittle-ductile transition and its geological implications. Can. Mineral 37:525-541 Wang ZC, Karato S, Fujino K (1993) High temperature creep of single crystal strontium titanate (SrTiO3)—a contribution to creep systematics in perovskites. Phys Earth Planet Int 79:299-312 Warren BE, Avenbach BL (1950) The effect of cold work distorsion on X-ray pattern. J Appl Phys 21: 595-610 Weidner DJ, Chen JH, Xu YQ, Wu YJ, Vaughan MT, Li L (2001) Subduction zone rheology. Phys Earth Planet Int 127:67-81 Wenk HR, Matthies S, Hemley RJ, Mao HK, Shu J (2000) The plastic deformation of iron at pressures of the Earth's inner core. Nature 405:1044-1047 Williamson GK, Hall WH (1953) X-ray line broadening from filed aluminium and wolfram. Acta Metall 1: 22-31 Wright K, Price GD, Poirier JP (1992) High-temperature creep of the perovskites CaTiO3 and NaNbO3. Phys Earth Planet Int 74:9-22

180

6

Cordier

Dislocations and Slip Systems of Mantle Minerals Patrick Cordier

INTRODUCTION....................................................................................................................... 137 BASIC CONSIDERATIONS...................................................................................................... 137 EXPERIMENTAL ADVANCES................................................................................................ 142 High-pressure deformation experiments............................................................................... 143 Transmission electron microscopy ....................................................................................... 144 X-ray diffraction peak broadening........................................................................................ 147 PLASTICITY OF MANTLE PHASES....................................................................................... 147 SiO2 system........................................................................................................................... 149 Olivine, wadsleyite, ringwoodite .......................................................................................... 153 Garnets .................................................................................................................................. 167 WHERE DO WE STAND?......................................................................................................... 171 PERSPECTIVES: FROM ATOMIC TO THE GLOBAL SCALE............................................. 173 REFERENCES............................................................................................................................ 174

7

Instability of Deformation Harry W. Green, II Department of Earth Sciences and Institute of Geophysics and Planetary Physics University of California Riverside, California 92521

Chris Marone Department of Geosciences Pennsylvania State University University Park, Pennsylvania 16802

INTRODUCTION This chapter addresses the mechanisms by which instabilities can develop in rocks in the nominally ductile regime under conditions that the authors believe are relevant to Earth. We will not address brittle failure or frictional sliding processes in detail, but much of our discussion will be couched in terms of knowledge gained in studies of these processes. As a consequence, we will begin by discussing why brittle failure and frictional sliding cannot be the mechanism by which rocks become unstable at depth, despite the strong evidence that they are the underlying mechanism of earthquakes at shallow depths (e.g., Scholz 1990, 2002). Brittle shear failure in the normal sense is fundamentally a tensile process (tensile microcracks must first nucleate and self-organize via interaction of the stress concentrations at their tips). As a consequence, brittle failure and frictional sliding are strongly inhibited by pressure because work must be done against the pressure to open the Mode I cracks. The inhibition is so strong that the stress required to create a fault or initiate sliding on an existing fault becomes greater than the room-temperature flow stress of many rocks at pressures equivalent to only a few tens of kilometers in Earth. Increasing temperature has little effect on brittle processes, but in contrast, the ductile flow stress of rocks falls exponentially with temperature. Because both pressure and temperature increase with depth in Earth, earthquakes by unassisted brittle fracture mechanisms or frictional sliding can only occur at depths less than ~30-50 km. Nevertheless, earthquakes occur to depths of almost 700 km in Earth and seismological evidence demonstrates unequivocally that they occur by displacement across a surface or narrow zone (i.e., they occur by faulting). This conundrum can be explained qualitatively by certain specific types of mineral reactions occurring under certain restricted conditions. The specific conditions required for faulting or sliding involve both the nature and kinetics of the mineral reactions and the nature of the way the rock that fails is loaded. That is, in order for a shearing instability to develop, certain rigorous conditions must be met involving the relative properties of the material that is faulting and its surroundings. In this chapter, we will outline the requirements for a shearing instability to develop, followed by discussion of two different self-organizing processes associated with three different categories of mineral reactions that are currently known to lead to faulting at pressures where conventional brittle fracture is inhibited. In addition, we will discuss the popular theoretical concept of thermal runaway in shear zones as an earthquake mechanism in the context of the nature of loading in Earth and the lack of any obvious self-organizing aspect. We will show that the three known reaction-related, self1529-6466/00/0051-0007$05.00

182

Green & Marone

organizing, faulting mechanisms have been characterized to various degrees and probably all have relevance to faulting in Earth. However, none of them has been sufficiently investigated to allow quantitative analysis of the criteria for instability in the same way that brittle deformation has been analyzed. SHEARING INSTABILITY As we have seen in the foregoing discussion, the great pressures at depth mean that earthquakes under such conditions cannot represent conventional frictional shear or fracture. Nevertheless, seismic studies show that deep earthquakes have double-couple focal mechanisms and moment tensors that are consistent with shear rupture along a narrow zone (Wiens et al. 1993, 1994). Thus, it is reasonable to pursue parallels between deep earthquake instability and shallow crustal earthquakes, which are comparatively well understood (e.g., Scholz 1990, 2002). By definition, all earthquakes radiate energy at seismic frequencies, and thus rupture must propagate at velocities approaching the elastic wave speed of the surrounding rock. Dynamic rupture velocities and stress drops of earthquakes at depth are interpreted to be in the range of 2 to 4.5 km/s and 10-100 MPa, respectively (Fukao, and Kikuchi 1987, Wiens and McGuire 1995, McGuire et al. 1997). These observations provide two important constraints on the mechanism of intermediate and deep earthquakes. First, they show that elastic strains must exist in the hypocentral region at great depths. Because ongoing plastic deformation during subduction would relieve elastic strains generated at shallow depths, the elastic strains that drive intermediate and deep earthquakes must result from local, active deformation. A second point is that these earthquakes represent a mechanical shear instability with dominantly shear components of motion. Thus, a key question is that of how the earthquakes nucleate and how strain rates increase from background levels of 10-15 to 10-12 s-1 to the values of 10-1 s-1 or more that are appropriate for seismic slip rates of 1-10 m/s. Intermediate and deep earthquakes represent mechanical shear instability in the sense that some form of sudden weakening causes strength to drop below the applied stress level, resulting in a local force imbalance and dynamic acceleration. Shear instability is the result of interaction between fault zone rheology and continuum coupling with the surroundings. There are two requirements for instability. First, the rheologic response must result in weakening; a perturbation in strain rate that results in strengthening will not grow to become dynamically unstable. Second, the rate of weakening must satisfy a stability criterion governed by the continuum interactions between the nucleation region and its surroundings. In the most general case of a gradual transition from quasi-static to dynamic slip, a complete model of shearing instability requires an understanding of both the initial weakening event and how its growth leads to degradation of the local elastic stiffness (e.g., Rice and Ruina 1983). Several mechanisms have been proposed for earthquakes at depth and the exact form of initial weakening is under investigation. Nevertheless, for our purposes we can discuss instability in a generic form and examine the two primary effects. We may start by considering a nucleation zone embedded in a homogeneous medium (Fig. 1). In terms of the mineral reaction mechanisms discussed below, the nucleation zone may be thought of as a single fluid-containing shear crack or a region containing a critical spatial-density of fluid-filled Mode I cracks (or comparable anticrack structures produced by phase transformation). At this stage, we do not consider shear heating or thermal runaway. Although one could envision Figure 1a as a region in which shear heating had reached a critical threshold via localized slip, we focus attention here on the process(es) that can lead to such initial slip localization. We assume a region of

Instability of Deformation

183

finite thickness W, and for simplicity, we take a circular region of radius r. We require only that the nucleation zone be mechanically weaker than the surroundings, so that its growth leads to weakening of the overall region, with stress concentration at the zone ends. The relationship between slip and weakening can be estimated using elastic dislocation theory: Gu Δτ = C , (1) r

where Δτ is the change in stress associated with slip u on a circular dislocation of radius r, G is the elastic shear modulus, and C is a constant. The effective elastic stiffness K of the region is then K = C

G . r

(2)

Earthquake instability requires that fault zone strength fall faster than the unloading stiffness K of the surrounding region. Figure 1b is drawn in a way that emphasizes slip weakening, however equivalent plots could be constructed for strain-rate weakening, slip velocity weakening, or phase-transformation induced weakening. The energy available to drive shear rupture (including that to overcome shear resistance, create surface area, radiate seismic energy, and cause heating) is given by the shaded region in Figure 1b.

Figure 1. (a) Schematic view of the nucleation zone for a generic model of deep earthquake instability. Stippled area shows weakened region of radius r and width W. Internal shear strain is given by g = u/W, where u is slip measured from one side to the other. (b) Stress-displacement plot for the nucleation zone and surroundings. Dashed line of slope K shows the unloading stiffness (see Eqn. 2) for the region surrounding the nucleation zone. Heavy line shows hypothetical strength curve for the nucleation zone. Note that initial weakening causes strength to fall below the applied stress, leading to a dynamic force imbalance. Shaded region shows energy available to drive seismic rupture.

The instability condition is determined by the relative rates of weakening and elastic stress release. This can be quantified with reference to a generic slip-weakening model, in which we assume linear weakening over a distance D (Fig. 2). Taking τ y as the yield strength and τ f as the failure strength, we can write the instability condition as: K < Kc ,

(3)

where Kc is a critical stiffness given by: Kc =

(τ

y

−τ

D

f

).

(4)

184

Green & Marone

Because the elastic stiffness is always positive, we see that weakening is a necessary condition for instability. We may define the following cases. τ y < τ f , Kc < 0 (5) Stable: Conditionally Stable

τy > τ f,

K > Kc

(6)

Unstable

τy > τ f,

K < Kc

(7)

Figure 2. Generic slip weakening model for earthquake instability. Fault strength (solid line) weakens from t y to t f over slip distance D or shear strain gc = D/W. Applied stress (dashed line) follows elastic unloading stiffness. Instability and seismic rupture are produced by the force imbalance between applied stress and fault strength.

In the first case, deformation is inherently stable because possible nucleation events, including slip or strain rate perturbations lead to strengthening. This is the standard expectation for steady-state plastic rheologies or viscous deformation in which strength increases with strain rate. The necessary condition for instability is met for the conditionally stable case, but the rate of weakening is insufficient to satisfy the instability criterion. In this case, the fault zone weakens but shear is stable and does not generate seismic radiation. In the final case (Eqn. 7), both the necessary and sufficient conditions are met and perturbations of any size could grow to become unstable. The dynamic force imbalance in this case would lead to rapid fault acceleration, shear heating and seismic radiation during slip. In the context of a stability analysis for earthquake nucleation at high pressure, a key question is that of how rapidly a fault weakens during initial slip. For the analysis presented here we use only slip weakening, hence we are assuming that weakening is a constitutive property of a fault zone, independent of strain rate, slip or other variables. To the extent that this approximation holds, the critical weakening rate for instability can be quantified as a rheologic stiffness, as done in Equation (4). By combining the relations for effective elastic stiffness (2) and the critical rheologic stiffness (4) we may write the stability criterion as: G (τ y − τ f ). < (8) r D This relation brings out an important point regarding conditionally stable shear (Fig. 3). In this case, the rate of weakening with slip is slower than the rate of stress reduction by unloading and thus shear is stable. However, growth of the nucleation region will reduce the local stiffness, which may eventually lead to instability. In the context of this approach, a critical nucleation patch size rc can be defined: C

rc = C

(τ

GD −τ f

y

.

)

(9)

Figure 3 shows that as the nucleation dimension grows from r1 to r3, the effective stiffness falls below the critical stiffness and the instability condition is satisfied.

Instability of Deformation

185

Figure 3. Slip weakening model of fault strength (solid line) together with the effective stiffness for three fault nucleation sizes (r1 < r2 < r3). Stiffness decreases with the nucleation dimension. Kc denotes critical stiffness defined by fault strength and rheology. A critical nucleation size for instability rc can be defined in terms of Kc.

In summary, earthquake instability requires that fault zone strength fall faster during weakening than the unloading stiffness of the surrounding region. In the context of a simplified, generic model of fault rheology, a critical stiffness for instability can be defined. A key question is that of how rapidly a fault weakens during initial slip. Clearly, much of the above analysis is highly oversimplified. More complex approaches would account for the effect of perturbation size on yield strength and failure conditions and incorporate specific mechanisms of plastic instability into relations for the critical rheologic stiffness. Also, the effect of latent heat of transformation, shear heating, and complex nucleation geometry would be included in the rheologic and continuum models. Quantitative treatment of these important effects is beyond the scope of this chapter and requires additional information about potential fault nucleation EXPERIMENTAL HIGH-PRESSURE FAULTING MECHANISMS Dehydration-induced embrittlement Breakdown of hydrous minerals. Raleigh and Paterson (1965; see also Raleigh 1967) conducted deformation experiments on serpentinite under confining pressures of a few hundred MPa. They found that at elevated temperatures the rock was ductile. However, under conditions where the serpentine was breaking down to olivine + talc + H2O, the rock failed catastrophically by faulting. This effect has been interpreted classically as a result of reduction of the effective normal stress on a fault plane or potential fault plane, resulting in the following relationship: τ = τo + µ(σn – pf),

(10)

where τ is the shear stress, τo is a constant, µ is the coefficient of friction, σn is the normal stress on the fault, and pf is the pressure in the pore fluid. Experimentation under conditions where the pore fluid pressure can be manipulated independently of the stress applied to the solid rock has verified this effective stress relationship for both brittle failure of intact rock and frictional sliding under conditions where pf < σ3. However, the qualitative description of the fluid as partially supporting the normal stress is unsatisfactory in a mechanistic sense because if pf < σ3, then the fluid can only exist in pores for which the strength of the solid matrix shelters the pores. If that is the case, how does the fluid partially reduce the normal stress on a fault? Worse, how does it partially reduce the normal stress on a fault that isn’t there yet? Clearly, this relationship may provide a satisfactory continuum description of the phenomenon, but not a mechanistic explanation. The answer comes from modern understanding that brittle shear failure is at its base

186

Green & Marone

tensile failure. Experimentally, it has been known for 40 years that small shear cracks cannot be made to propagate in their own plane. Application of a stress field that would be expected to make such a crack propagate causes “wing” cracks to originate at sites of greatest tensile stress concentration at the tips of the crack and propagate along the trajectory of σ1, coming to rest when they are sufficiently far from the stress concentration that initiated them for the stress at the crack-tip to fall below the local tensile fracture strength. Thus, macroscopic shear failure initiates by a gradual process of local tensile failure at points of stress concentration (pre-existing cracks, grain boundaries, pores, grains of contrasting elastic properties, etc.), culminating in a critical population of tensile microcracks that then self-organize to initiate a shear fracture (a fault) that propagates by repetition of the “microcracks-first” process along the path of the growing fault surface. Although perhaps counterintuitive and an unwelcome complication, this insight provides a built-in understanding of how a pressurized pore fluid facilitates faulting. It does so by enhancing the local tensile stress concentrations that lead to formation of the tiny tensile microcracks and by holding them open after they form (Fig. 4). Thus, the greater the pore pressure, the lower the applied stress necessary to bring local stress concentrations to the failure stress, and the lower the overall stress must climb to in order to generate proliferation of microcracks to the point of self-organization into an incipient fault zone that can then grow by the same pore-pressure-induced process. If the pore pressure should reach σ3, the expected failure mode would be runaway propagation of a tensile crack, leading to sample splitting in the laboratory or a fluid-filled vein (or meltfilled dike) in Earth. In the laboratory, this is easily demonstrated to be the case. However, in a natural situation with a slowly evolving pore pressure, the more usual case should be generation of shear failure before the pore pressure reaches σ3.

Figure 4. Fluid-assisted opening of “tensile microcracks.” Nucleation of a droplet of lowdensity fluid (net ΔV of reaction greater than zero) at a stress concentration in a nonhydrostatic stress field induces deviatoric tensile stresses around the bubble, with maximum tangent to the plane normal to σ1 as shown on left. Such stresses will enhance nucleation of a tensile microcrack that will grow until the stress at the crack tip produced by the combination of the applied stress and the fluid pressure within the crack falls below the local tensile fracture stress.

There is abundant laboratory evidence at moderate pressures that this mechanism can induce brittle failure at greatly reduced stresses compared to those required in the absence of a pore fluid. Importantly, as was the case in the experiments of Raleigh and Paterson

Instability of Deformation

187

(1965), development of a pore pressure can lead to brittle failure where otherwise (i.e., in the absence of a fluid) rock behavior will be ductile. As a consequence, dehydration embrittlement is an attractive hypothesis to explain earthquakes at pressures where unassisted brittle failure is completely inhibited. There is a problem, however. Fluids are very much more compressible than solids, hence dehydration reactions, which all produce a positive ΔV of reaction at low pressures (because of the very low density of hydrous fluids), will see the magnitude of ΔV progressively reduced as pressure rises and ΔV will become negative at some point (because the net effect of the more dense solid product phases will come to dominate). For antigorite, the serpentine mineral stable at the highest temperatures and pressures, that point is at ~750°C, 2.2 GPa (Ulmer and Trommsdorff 1995), a pressure representing a depth of only ~70 km in Earth (Fig. 5). For other hydrous minerals likely to be reasonably abundant in mantle lithologies, the conditions are comparable.

Pressure (GPa)

Maximum watercarrying geotherm

Negative slope implies negative volume change

Figure 5. Antigorite phase diagram. Decomposition relations for a natural antigorite (modified after Ulmer and Trommsdorff 1995). Between ~2.2 and 6.4 GPa, dehydration yields anhydrous phases and the slope of the dehydration boundary is negative, implying ΔV < 0. Above 6.4 GPa, Phase A, a dense hydrous magnesium silicate, forms rather than olivine and the slope of the dehydration curve becomes more negative.

Temperature, °C

However, if ΔV < 0, the conventional rock mechanics view is that the pore-pressure effect will disappear (e.g., Wong et al. 1997). That is because when the net ΔV becomes negative, generation of a droplet of fluid results in a low-pressure pore surrounded by compressive hoop stresses, hence any stress concentrations that encourage formation of Mode I microcracks are pushed away from local failure rather than toward it. The greater the negative ΔV, the lower is the pore pressure produced and the greater the compressive “capsule” in which it is confined. Both effects push the material away from local tensile failure, hence this has been interpreted as inhibiting bulk shear failure. If the latter interpretation is correct, then dehydration embrittlement of serpentine and other common hydrous alteration minerals of peridotite (e.g., chlorite) will be limited to relatively shallow depths as a mechanism for triggering earthquakes. This hypothesis has been tested in the laboratory (Reinen et al. 1998) and found that dehydration of antigorite under stress at confining pressures above 3 GPa still leads to

188

Green & Marone

faulting, showing that bulk shear failure can still occur with ΔV < 0. However, in keeping with the predictions of brittle rock mechanics, the microstructures accompanying faulting at those high pressures show no evidence of the Mode I microcracks diagnostic of brittle shear failure and fault gouges (Fig. 6) also do not consist of angular fragments with a fractal size distribution. These observations indicate that the mode of failure is not traditional frictional failure nor embrittlement of that type assisted by dehydration.

Figure 6. Faulting of antigorite when ΔV < 0. Despite the fact that ΔV for the dehydration reaction becomes negative above 2.2 GPa, dehydration under stress leads to faulting. How-ever, neither Mode I microcracks nor a fault gouge with angular fragments is observed.

Much remains to be learned about the fundamental physics of failure under these conditions. Nevertheless, at least we know that dehydration of hydrous minerals remains a viable trigger mechanism for earthquakes to depths of at least 250 km and probably much more, regardless of the sign of the ΔV of dehydration (so long as hydrous phases are present and are breaking down to form less-hydrous assemblages). Exsolution of H2O from nominally anhydrous minerals. A new variant of dehydration embrittlement recently has been discovered that provides additional insight into fluid-enhanced faulting (Green 2001). In a study of the rheology of eclogite at P = 3 GPa, experiments on a “wet” eclogite resulted in faulting when specimens were deformed at temperatures between the wet and dry solidi. The experimental specimens were fabricated from a natural eclogite with no hydrous phases, but with significant H2O dissolved in both of the major, nominally anhydrous, phases (omphacitic pyroxene and garnet). Above the water-saturated solidus, H2O exsolved from the silicates and triggered melting at grain boundaries. When this happened under stress, the specimen failed by faulting and was characterized by myriads of glass-filled Mode I cracks as well as a “normal” fault gouge except for the presence of small amounts of glass. In this case, ΔV was clearly greater than zero because the only phases being produced were fluid and melt, both of which are less dense than either omphacite or garnet; the microstructural characteristics of the specimens are as expected for ΔV > 0. Thus, this failure was truly dehydration embrittlement in the original sense. Change of micromechanism of dehydration-induced failure with change in sign of ΔV. The change of microstructures associated with dehydration-induced failure when the sign of net ΔV changes from positive to negative strongly suggests a change in failure mechanism. We defer discussion of this potential mechanism change until after discussion of transformation-induced faulting. Transformation-induced faulting The other experimentally established high-pressure faulting instability is the anticrack mechanism known to operate under certain restrictive conditions during the olivine → wadsleyite and olivine → ringwoodite transformations (Green and Burnley 1989,

Instability of Deformation

189

σ1 Dense Nucleus

σ3

σ3 Nanocrystal -filled Anticrack

σ1 Figure 7. Nucleation and growth of anticracks is analogous to nucleation and growth of fluid-filled cracks during dehydration under stress (compare with Fig. 4). In this case, formation of a nucleus of the dense polymorph induces strong compressive stress concentrations and a small rise in T that, in turn, cause rapid-fire nucleation in the plane normal to s1. The result is microanticracks that have high compressive stresses at their tips and are filled with a nanocrystalline aggregate of the denser phase that has fluid-like properties even at seismic slip rates.

Green et al. 1990, Burnley et al. 1991, Green & Houston 1995, Kirby et al. 1996). This is also a self-organizing instability; it requires an exothermic polymorphic phase transformation that leads to production of Mode I microanticracks (Fig. 7) filled with an extremely fine-grained aggregate of the new phase that interact via the compressive stress concentrations at their tips in an analogous way to the interactions between tensile stress concentrations at the tips of open or fluid-filled Mode I microcracks (Green and Burnley 1989). The reason an exothermic polymorphic transformation is required to support this instability is shown in Figure 8. Cooling the high-temperature phase below the phase boundary causes it to become metastable; as the difference in Gibbs Free Energy (ΔG) between the two phases increases, so does the nucleation rate. However, when undercooling is sufficiently great, the nucleation rate becomes more dependent on the kinetics of the transformation than on the driving force, hence the nucleation rate reaches a maximum and at lower temperatures it declines at an increasingly rapid rate; at temperatures lower than defined by the left-hand branch of the curve, the reaction rate is zero. Along the near-vertical left-hand branch of the nucleation-rate curve, formation of a nucleus of the stable phase will yield a small heat release that will result in a small increase in local temperature, leading to increase in the nucleation rate. At the same time, if ΔV of the reaction is negative, the nucleus immediately will be surrounded by compressive stress concentrations that will increase the driving force for nucleation. If the material is under stress, the maximum compressive stress concentration around the initial nucleus will be located adjacent to the nucleus on planes oriented normal to σ1, hence additional nuclei will preferentially form adjacent to the first nucleus and lie in the plane normal to σ1. This anisotropic contraction of the transformed region increases the stress concentrations lying in the plane normal to σ1. Therefore, if the thermal conduc-tivity of the metastable host is low and the kinetics of plastic relaxation of the stress concentrations are sufficiently slow, the combination of these two effects can lead to runaway nucleation in the plane normal to σ1 – yielding a Mode I anticrack filled with a nanocrystalline solid with properties approximating that of a low-

Green & Marone

Log Nucleation Rate

190

Figure 8. Thermodynamics of nucleation during polymorphic phase transformations. The topology of nucleation kinetics for transformation at constant pressure is described for exothermic reactions to the left of the vertical dashed line and for endothermic reactions to the right of that line. Examples are shown for cases where the highpressure phase is the low-entropy phase (left) and the high-entropy phase (right). See text for discussion (modified after Green and Zhou 1996).

viscosity fluid. As in the case with Mode I cracks, the greater the aspect ratio of the anticrack, the greater the stress concentration at the tip. Therefore, microanticracks interact with each other in a stressed material in an analogous way to microcracks, culminating in initiation of a fault and catastrophic shear failure. In contrast, endothermic transformations have entirely different characteristics (right side of Fig. 8). As the low-temperature polymorph is progressively heated above the transformation temperature, both the driving force and the kinetics for the transformation increase monotonically. As a consequence, there is no kinetically controlled branch of the nucleation-rate curve and any thermal effects of nucleation would actually produce a small reduction of local temperature. Thus, the characteristics that can lead to positive feedback are absent; when the temperature is sufficiently high for the reaction to run, simple nucleation and growth of the high-temperature phase occurs, leading to growth of normal, blocky, crystals; no anticracks form and no instability is generated. This argument strongly suggests that the sign of ΔS (≥ ΔH/T) is critical to this instability; if the reaction is not exothermic, there will be no instability. However, nothing in the logic presented here requires that the algebraic sign of ΔV should matter. The magnitude of ΔV controls the magnitude of the stress concentration at the tips of microlenses, but there is no reason that the sign of ΔV should matter. This implication was tested in the CdTiO3 system (Green and Zhou 1996); in the up-temperature (endothermic) direction, no instability developed. Under all conditions, transformation from the phase with ilmenite structure to that with perovskite structure (ΔV < 0; ΔH > 0) yielded blocky crystals of the daughter (high-temperature) phase growing on grain

Instability of Deformation

191

boundaries of the parent phase. In contrast, as described in the previous paragraph, when tested in the reverse direction (ΔV > 0; ΔH < 0), the reaction produced fine-grained lenses (“microcracks”) of the ilmenite phase parallel to σ1 and shear fracture followed. The logic for runaway growth of nanocrystalline lenses when ΔV > 0 is analogous to that given above for the case ΔV < 0; in this case the first nucleus is surrounded by extensile hoop stresses that prejudice the location of succeeding nuclei and the small amount of heat released leads to runaway nucleation in a plane parallel to σ1. Clearly, then, a critical requirement for failure is an exothermic reaction. But is that a sufficient cause? The reaction ringwoodite → perovskite + magnesiowüstite that defines the base of the upper mantle is endothermic but if olivine were to be carried metastably completely through the transition zone and then decompose into the lower-mantle assemblage, the reaction would be exothermic. It is necessary, therefore, to know whether such a disproportionation reaction could support the instability if the reaction is exothermic and ΔV is large. The reason to question this possibility is that a diffusive step is required for phase separation during this type of reaction. Is the time necessary to separate the chemical components into two phases sufficiently long that it abrogates the localized runaway nucleation of the daughter assemblage to form anticracks and lead to instability? This hypothesis was tested using the reaction albite → jadeite + coesite , because it is both more exothermic and has a larger volume change than olivine → spinel in Mg2GeO4, the system in which we discovered transformation-induced faulting (Green and Burnley 1989), or in the olivine → wadsleyite transformation in olivine of mantle composition (Green et al. 1990). The hypothesis was confirmed; decomposition of albite under stress yielded symplectites consisting of blocky crystals of jadeite filled with wormy intergrowths of coesite; no anticracks were observed and faulting did not occur (Gleason & Green 1996). Table 1 summarizes the systems in which the anticrack mechanism has been investigated, verifying that it can operate during a variety of exothermic polymorphic transformations but not during endothermic transformations nor during disproportionation reactions, even when ΔV is large and the heat evolved is large and positive. Table 1. Systems tested for transformation-induced faulting. Reaction

System

Faulting? Pressure (Gpa) Microstructure Reaction type

α→γ

Mg2GeO4

Yes

1-2

anticracks

α→β

(Mg,Fe)2SiO4

Yes

14-15

“

“

“

“

α→β

Mn2GeO4

Yes

4-4.5

“

“

“

“

pv →il

CdTiO3

Yes

0.2

“cracks”

“

“

“

Yes

0.2-0.5

“anticracks”

“

“

“

No

0.2

Blocky xtls

Endo. “

“

No

3-3.5

Symplectite

Exo. disproportion.

Ice1h → II* H2O il → pv

CdTiO3

ab →jd + coes

Exo. poly. trans

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Predictions for the top of the lower mantle: γ → pv + mw

No

25

Symplectite Endo. disproportion.

α → pv + mw

No

25

Symplectite Exo. disproportion.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ *Kirby et al. (1991); compression features similar to anticracks were observed.

192

Green & Marone

Figure 9. Comparison between low-pressure brittle failure of Westerly Granite (left) and highpressure anticrack failure of Mg2GeO4 undergoing the olivine→spinel transformation. Note that brittle failure produces an exponentially increasing number of acoustic emissions that rises very high before failure, whereas anticrack failure produces significant acoustic emissions only during macroscopic failure (modified after Green, 1994).

Like normal brittle failure, anticrack failure in the laboratory generates acoustic emissions (Green et al. 1992) and thus is potentially an earthquake mechanism (Fig. 9). “Brittle” versus “plastic” shear failure

Transformation-induced failure clearly operates by a different physical process than conventional brittle failure, yet it mimics many aspects. In particular, it achieves selforganization via Mode I nanocrystalline lenses (either parallel or perpendicular to σ1, depending on the sign of ΔV. Nevertheless, it produces acoustic emissions only during shear failure; the anticrack growth phase before bulk failure is silent (Fig. 9). This observation plus the current interpretation of the underlying physics of anticrack growth as a runaway nucleation phenomenon indicates that this failure mechanism is perhaps better classified as a plastic instability rather than brittle failure. For dehydration-induced faulting, the change in microstructures associated with change in sign of ΔV also suggests a change in failure mechanism. In particular, change from Mode I microcracks and angular, fractal, fault gouge to microstructures indicative of mineral reactions suggests that dehydration with negative ΔV also may trigger a plastic instability. The details of this proposed switch of mechanisms remains unclear. Thermal runaway due to shear heating

In addition to dehydration-induced faulting and transformation-induced faulting, shear failure mechanisms that have been established in the laboratory, it has been proposed a number of times on theoretical grounds that a shearing instability can develop based on thermal feedback between shear heating and thermally-induced viscosity reduction under conditions approximating constant stress (e.g., Gruntfest 1963, Griggs and Baker 1969, Hobbs and Ord 1988, Kanamori et al. 1998, Karato et al. 2001). These treatments leave little doubt that such a runaway could lead to failure. What is not clear is whether boundary conditions in Earth are such that the runaway can be supported. For

Instability of Deformation

193

example, under constant strain-rate or constant displacement conditions, the thermallyinduced viscosity resulting from shear heating will lead to stress relaxation rather than runaway heating, and any thermal anomaly will dissipate rather than leading to failure. As described above, the anticrack mechanism also appears to be a thermal runaway phenomenon. Why is this mechanism self-organizing and shear heating not? Like in brittle failure, the anticrack instability is in the development of the primary (Mode I) failure mechanism. The thermal runaway in anticrack development works off of latent heat release and therefore is proportional to reaction progress and depends on the relative rates of thermal conductivity and reaction. No energy need be supplied externally during the heating process and the volume of material transformed before failure can be extremely small. In contrast, shear heating is proportional to the work done during straining and therefore is a function of stress, strain, viscosity, thermal conductivity, and geometrical boundary conditions (the system must be pre-organized into an appropriately-oriented shear zone). Instability can be obtained only if the boundary conditions ensure that the viscosity reduction induced by increasing temperature results in increase of strain rate rather than relaxation of stress. APPLICATION TO EARTHQUAKE MECHANISMS Earthquake distribution with depth

EARTHQUAKES PER YEAR

Figure 10 shows the distribution of earthquakes in Earth. Virtually all earthquakes deeper than 20-30 km are generated at the interface between subducting lithosphere and the over-riding plate, or within the subducting lithosphere itself, hence this distribution shows that the generation of earthquakes with depth is generally spread out over the entire depth of subduction to the base of the mantle transition zone, where they stop abruptly. Thus, it is to be expected that a successful candidate for an earthquake trigger mechanism will naturally produce earthquakes continuously with depth, rather than bursts at particular set(s) of conditions.

Figure 10. Semi-log plot of earthquake frequency with depth. Note exponential decline from 100 to 300 km and secondary population in mantle transition zone reaching a maximum at ~600 km before total cessation at 680 km (modified after Frohlich 1989).

194

Green & Marone

Mineral reactions available to trigger earthquakes

We know from dredge hauls and heat flow that the upper portion of the oceanic lithosphere undergoes hydrothermal alteration shortly after its formation at oceanic ridges. Seismic velocities suggest that such alteration extends into the mantle beneath the crust and therefore that serpentine partially replaces olivine (and pyroxene), perhaps up to a depth of 10-12 km (Meade and Jeanloz 1991). The rare evidence for hydrous alteration in shallow mantle xenoliths from oceanic islands strongly supports this view. Thus, it has been generally assumed (e.g., Green and Houston 1995) that dehydration-induced faulting is an excellent candidate for triggering these earthquakes, but that this mechanism is probably not a candidate to explain the common occurrence of a second seismic zone deep within the lithosphere at intermediate depths (e.g., 40-200 km below Japan, Igarashi et al. 2001). However, Silver et al. (1995; see also Jiao et al. 2000) suggested that deeper hydration of the oceanic lithosphere might occur along fault zones created by great earthquakes outboard of trenches and others have pointed out the similar pattern of maximum antigorite stability in subduction zones (e.g., Ulmer and Trommsdorff 1995) and the distribution of earthquakes in intermediate-depth double seismic zones (e.g., Peacock 2001). More recently, a somewhat different mechanism has been proposed that suggests extensive serpentinization may occur at trenches (PhippsMorgan 2001, Phipps-Morgan et al. 2002). If it is true that the subducting lithosphere is hydrated at least sporadically to a depth of 40 km, dehydration-induced failure is a potential candidate for explanation of all earthquakes at depth. Here we will briefly examine this possibility in three depth intervals: 0-200 km, 200-300 km, 300-700 km.

Figure 11. Stability of antigorite mapped onto a thermal model of a subduction zone. The particular model is such that antigorite stability reaches that of Phase A (~200 km depth). Hydrous phases existing in the cold core of the slab could access the antigorite→Phase A + en reaction and pass H2O into the deeper slab. Otherwise, all H2O will be released and could flux arc melting (heavy arrow) or trigger earthquakes, but could not be carried to great depths in significant quantities. Note similarity between limit of antigorite stability (heavy line) and earthquake distribution beneath northern Japan (inset). Hachured region represents cold lithosphere. (Modified after Ulmer and Trommsdorff 1995 and Igarashi et al. 2001).

Instability of Deformation

195

(i) 0 - 200 km. Depending on the thermal profile of a subduction zone, antigorite can remain stable in the cold interior of the descending slab up to a maximum depth of approximately 200 km. As shown in Figure 11, the shape of the antigorite phase diagram in the thermal model chosen is remarkably similar to that of the earthquake distribution under northern Japan. As a consequence, dehydration of antigorite could explain the entire distribution of earthquakes in the deeper zone, including its approach to and merging with the shallower zone. In addition, antigorite, in combination with a myriad of other hydrous phases potentially present in the hydrated oceanic crust and shallow mantle, can also explain the shallow zone. Of course, the specific depth to which antigorite could carry water and release it to trigger earthquakes is a function of the temperature distribution in the slab. As was shown in Figure 5, the temperature of maximum stability of antigorite falls rapidly at pressures greater than ~7 GPa, hence for water to pass to greater depths, it must be passed on to hydrous phases stable to greater depths. In particular, if the minimum temperature in the slab reaches ~550-600°C before a depth of ~200 km is reached, antigorite will completely dehydrate to olivine + enstatite and the slab will be completely dehydrated except for a few rare hydrous phases. However, Figure 5 also shows that if the slab is colder, antigorite will break down into Phase A + H2O and the door will be open to carry water into the deep interior via a succession of dense hydrous magnesium silicate phases (e.g., Angel et al. 2001). Indeed, this “eye of the needle” has profound importance for mantle dynamics and therefore for past, present, and future evolution of Earth. (ii) 200-300 km. Deeper than 200 km, a problem arises for earthquake generation by dehydration. Examination of the data of Angel and Frost (2001) shows that any reasonable particle trajectory within a cold slab (a slab for which the minimum temperature passes below ~600°C, 7 GPa) crosses all of the phase boundaries available in directions for which water is conserved. That is, each hydrous phase contains less water than the more dense hydrous phase(s) that replace it. Thus, no free fluid is generated by the reactions and dehydration would appear to be unavailable for triggering earthquakes. There are, however, a few hydrous phases that would be expected to be present in minor quantities in hydrous subducting lithosphere that could potentially extend dehydration embrittlement to depths somewhat greater than 200 km. For example, titanian clinohumite is stable to depths exceeding 300 km (Ulmer and Trommsdorff 2001) and could perhaps explain the low frequency of earthquakes at that depth. (iii) 300-700 km. The previous paragraph suggests that dehydration cannot serve as a trigger for earthquakes at depths greater than 300-400 km. In contrast, restriction of the anticrack instability to exothermic polymorphic transformations predicts that earthquakes can only be triggered by this mechanism at depths where the transformations of olivine to wadsleyite and ringwoodite occur (i.e., at depths of 300-700 km in subducting slabs). Therefore, this mechanism could explain the rise in frequency of earthquakes at the top of the transition zone and their cessation at its base (Green and Houston 1995). However, application of this mechanism to Earth requires that the central portion of descending slabs is sufficiently cold that olivine does not transform to its denser polymorphs until a critical set of conditions is reached where the kinetics of the reactions are rapid enough to run. It also may seem that the numerous special circumstances necessary for anticrack failure would make the probability small that it would happen routinely in Earth. However, in subduction zones, a given volume of material follows a particle path that is primarily up-pressure, but nevertheless is also continually encountering conditions of higher temperature as heat flows slowly into the slab from the surrounding mantle. As a consequence, if conditions are sufficiently cold in the central parts of a subducting slab for reaction to be inhibited at the top of the transition zone, a wedge of metastable olivine is created, all volumes of which must eventually encounter the conditions critical for

196

Green & Marone

transformation-induced failure unless they pass into the lower mantle and transform there passively to perovskite + magnesiowüstite. Unfortunately, at the present time, neither the exact temperature distribution in slabs nor the exact conditions for metastable olivine preservation are known. Seismic evidence that such metastable olivine is present in subducting slabs traversing the transition zone has been equivocal (cf. Koper et al. 1998) and some calculations based upon experimental studies of the kinetics of the transformations and thermal models of subduction zones have suggested that metastable olivine is likely only in the coldest slabs (Mosenfelder et al. 2001). Recently, however, evidence has been collected showing slow seismic velocities in a subhorizontal, slab-like, earthquakegenerating anomaly beneath Fiji that is most easily explained as a remnant slab of metastable olivine floating in the transition zone (Chen and Brudzinski 2001, Green 2001). If this interpretation is valid, the conflict between observation and prediction may indicate that subduction zones are colder than suggested by current thermal models. However, it is premature to reach such a conclusion. In addition to the uncertainties in slab temperature distributions and olivine-spinel reaction kinetics, there are specific characteristics of certain earthquakes that are difficult to explain by either of these mechanisms. Some very deep earthquakes, most notably those in the South American slab, are interpreted from their seismic characteristics to have involved seismic slip zones of diameter several 10s of km. For example, both the 1994 Tonga and Bolivia deep earthquakes have been estimated to involve slip on regions with diameters of ~60 km. Therefore, it is difficult to explain the full slip in these earthquakes by any mechanism that is inherently restricted to cold temperatures (i.e., both transformation-induced faulting and dehydration-induced faulting). Therefore, even if slab temperatures are colder than presently envisioned, it remains questionable as to whether there could be sufficiently large volumes of metastable olivine or hydrated mantle available to support such slip. It is important, therefore, to consider possible alternative or cooperative mechanisms. It is this concern that has led to resurgence of thermal runaway models for the deepest earthquakes (e.g., Karato et al. 2001). CONCLUSIONS AND SPECULATIONS

The discussion presented here shows that dehydration-induced faulting and transformation-induced faulting are complementary; the former appears to be restricted to depths above ~350 km and the latter to depths below ~350 km. Moreover, the waning ability of dehydration to trigger earthquakes below 200 km conveniently falls in the region of lowest earthquake production and the inability of transformation-induced faulting to trigger faulting in endothermic or disproportionation reactions offers a natural explanation for cessation of earthquakes at the base of the transition zone. It is thus possible that earthquake initiation at depth in Earth can be completely explained by a combination of these two mechanisms. Nevertheless, it is premature to reach this conclusion for three principal reasons: (i) It is not clear that double seismic zones at intermediate depths can be assigned to dehydration-induced faulting; (ii) very large, very deep earthquakes seem to be inconsistent with expectations about the size and geometry of a metastable wedge of olivine; (iii) combination of experimentally-measured kinetics of olivine breakdown reactions and thermal models suggest that subduction zones in general are too warm for metastable olivine to persist to significant depths other than in

Instability of Deformation

197

Tonga, although if Chen and Brudzinski (2001) have truly observed metastable olivine in an older slab floating beneath Fiji, either the kinetics of transformation, the extrapolation of those kinetics to mantle conditions, or slab models may need revision. Additional work on this problem needed at this time includes (i)

laboratory experiments to determine whether the dense hydrous magnesium silicates are capable of generation of faulting via transient release of H2O during reaction from one to another;

(ii)

quantitative determination of the conditions of failure of olivine of mantle composition as a function of pressure and development of accurate methods to extrapolate those conditions to natural time scales;

(iii)

investigation of mechanisms of hydration of oceanic lithosphere at subduction zones;

(iv)

a thorough analysis of the errors potentially involved in experimental measurement of the kinetics of olivine reaction at high pressures and temperatures, their extrapolation to natural conditions, and the assumptions involved in thermal modeling;

(v)

evaluation of alternative mechanisms in terms of their ability to function as trigger mechanisms for earthquakes under natural boundary conditions. REFERENCES

Angel RJ, Frost DJ, Ross NL et al. (2001) Stabilities and equations of state of dense hydrous magnesium silicates Phys Earth Planet Inter 127:181-196 Antolik M, Dreger D, Romanowicz B (1996) Finite fault source study of the great 1994 deep Bolivia earthquake. Geophys Res Lett 23:1589-1592 Burnley PC, Green HW II, Prior D (1991) Faulting associated with the olivine to spinel transformation in Mg2GeO4 and its implications for deep-focus earthquakes. J Geophys Res 96:425-443 Chen WP, Brudzinski, M.R (2001) Evidence for large-scale, imbricate remnant of subducted lithosphere. Science 292:2475-2479 Frohlich C (1989) The nature of deep-focus earthquakes. Ann. Rev Earth Planet. Science 17:227-254 Fukao Y, Kikuchi M (1987) Source retrieval for mantle earthquakes by iterative deconvolution of longperiod P-waves. Tectonophysics 144:249-269 Gleason G, Green HW II (1996) Effect of differential stress on the albite to jadeite + coesite transition at confining pressures of >3 GPa. EOS, Trans Am Geophys Union 77:F662 Green HW II (1994) Solving the paradox of deep earthquakes. Sci Am 271:64-71 Green HW II (2001) Deep Tonga backarc: Graveyard for buoyant slabs bearing metastable olivine? Science 292:2445-2446. Green HW II (2001) Physical mechanisms for earthquakes at intermediate depths. EOS Trans Am Geophys Union 82, Fall Mtg Suppl, Abstr S42D-03 Green HW II, Burnley PC (1989) A new self-organizing mechanism for deep-focus earthquakes. Nature 341:733-737 Green HW II, Houston H (1995) The mechanics of deep earthquakes. Ann. Rev Earth and Planet. Sci. 23:169-213 Green HW II, Scholz CH, Tingle TN, Young TE, Koczynski T (1992) Acoustic emissions produced by anticrack faulting during the olivine → spinel transformation. Geophys Res Lett 19:789-792 Green HW II, Young TE, Walker D, Scholz CH (1990) Anticrack-associated faulting at very high pressure in natural olivine. Nature 348:720-722 Green HW II, Zhou Y (1996) Transformation-induced faulting requires an exothermic reaction and explains the termination of earthquakes at the base of the mantle transition zone. In Avé Lallemant et al. (eds) Carter Volume. Tectonophysics 256:39-56 Griggs DT, Baker DW (1969) The origin of deep-focus earthquakes. In Properties of Matter Under Unusual Conditions. H Mark, S Fernbach (eds) New York: Wiley Interscience, p 23-42

198

Green & Marone

Gruntfest IJ (1963) Thermal feedback in liquid flow—plane shear at constant stress. Trans Soc Rheology 7:195-207 Hobbs BE, Ord A (1988) Plastic instabilities: implications for the origin of intermediate and deep focus earthquakes. J Geophys Res 9323-42,10521-10540 Igarashi I, Matsuzawa T, Umino N, Hasegawa A (2001) Spatial distribution of focal mechanisms for interplate and intraplate earthquakes associated with the subducting Pacific plate beneath the northeastern Japan arc: A triple-planed deep seismic zone. J Geophys Res 106:2177-2191 Jiao W, Silver PG, Fei Y, Prewitt CT (2000) Do deep earthquakes occur on preexisting weak zones? An examination of the Tonga subduction zone. J Geophys Res 105:28125-38 Kanamori H, Anderson DL, Heaton TH (1998) Frictional melting during the rupture of the 1994 Bolivian earthquake. Science 279:839-842 Karato S, Riedel MR, Yuen DA (2001) Rheological structure and deformation of subducted slabs in the mantle transition zone: implications for mantle circulation and deep earthquakes. Phys Earth Planet Inter 127:83-108 Kirby SH, Durham WB, Stern L (1991) Mantle phase changes and deep earthquake faulting in subducting lithosphere. Science 252:216-25 Kirby SH, Stein S, Okal EA, Rubie DC (1996) Metastable mantle phase transformations and deep earthquakes in subducting oceanic lithosphere. Rev Geophys 34, 261-306 Koper KD et al. (1998) Modeling the Tonga Slab: can travel time data resolve a metastable olivine wedge? J Geophys Res 103:30079-30100 McGuire JJ, Wiens DA, Shore PJ, Bevis MG (1997) The March 9, 1994 deep Tonga earthquake: Rupture outside the seismically active slab. J Geophys Res 102:15163-15182. Meade C, Jeanloz R (1991) Deep-focus earthquakes and recycling of water into the earth's mantle. Science 252:68-72 Mosenfelder JL, Marton FC, Ross CR, et al. (2001) Experimental constraints on the depth of olivine metastability in subducting lithosphere. Phys. Earth Planet. Inter 127:165-180 Ogawa M (1987) Shear instability in a viscoelastic material as the cause of deep focus earthquakes. J Geophys Res 92:13,801-13,810 Peacock SM (2001) Are the lower planes of double seismic zones caused by serpentine dehydration in subducting oceanic mantle? Geology 29:299–302 Phipps-Morgan J (2001) The role of serpentinization and deserpentinization in bending and unbending the subducting slab. EOS, Trans Am Geophys Union 82, Fall Mtg Suppl, Abstr T22D-03 Phipps-Morgan J, Ranero C, Ruepke L (2002) Serpentinization, deserpentinization, and the bending and unbending of subducting slabs (abstr) Proc Structure and Tectonics of Convergent Plate Margins. Prague, Czech Republic (in press) Raleigh CB (1967) Tectonic implications of serpentinite weakening. Geophys J Royal Astron Soc 14: 45-51 Raleigh CB, Paterson MS (1965) Experimental deformation of serpentinite and its tectonic implications. J Geophys Res 70:3965-3985 Reinen LA, Green HW II, Nielsen SK (1998) Dehydration embrittlement of antigorite serpentinite at high pressures: Implication for intermediate depth earthquakes. EOS, Trans Am Geophys Union 79:F853 Rice JR, Ruina AL (1983) Stability of steady frictional slipping. J Appl Mech 50, 343-49 Scholz CH (1990) The Mechanics of Earthquakes and Faulting. Cambridge, UK: Cambridge University Press, 439 p Scholz CH (2002) The Mechanics of Earthquakes and Faulting, 2nd Edition. Cambridge, UK: Cambridge University Press, 496 p Silver PG, Beck SL, Wallace TC, Meade C, Myers S, James D, Kuehnel R (1995) The rupture characteristics of the deep Bolivian earthquake of 1994 and the mechanism of deep-focus earthquakes. Science 268:69-73 Ulmer P, Trommsdorff V (1995) Serpentine stability to mantle depths and subduction related magmatism. Science 268:858-861 Ulmer P, Trommsdorff V (1999) Phase relations of hydrous mantle subducting to 300 km. In Fei Y, Bertka CM, Mysen BO (eds) Mantle Petrology: Field Observations and High Pressure Experiments. Geochem Soc Spec Publ 6:259-281 Wiens DA, McGuire JJ (1995) The 1994 Bolivia and Tonga events: Fundamentally different types of deep earthquakes? Geophys Res Lett 22:2245-2248. Wiens DA, McGuire JJ, Shore PJ (1993) Evidence for transformational faulting from a deep double seismic zone in Tonga. Nature 364:790-793

Instability of Deformation

199

Wiens DA, McGuire JJ, Shore PJ, Bevis MG, Draunidalo K, Prasad G, Helus SP (1994) A deep earthquake aftershock sequence and implications for the rupture mechanism of deep earthquakes. Nature 372: 540-543 Wong T-F, Ko SC, Olgaard DL (1997) Generation and maintenance of pore pressure excess in a dehydrating system, 2. Theoretical analysis. J Geophys Res 102:841-852

8

Brittle Failure of Ice Erland M. Schulson Thayer School of Engineering Dartmouth College Hanover, New Hampshire 03755 Reprinted with modifications from Engineering Fracture Mechanics, vol. 68, p. 1839-1887, copyright 2001, with permission from Elsevier Science.

INTRODUCTION Ice exhibits either ductile or brittle behavior, depending upon the conditions under which it is loaded. Glaciers, for instance, are loaded by gravity, under deviatoric stresses of ~0.1 MPa or lower. At temperatures of interest, they flow through dislocation creep at strain rates of the order of 10-9 s-1 or lower (Patterson 1994). Sheets of sea ice—another terrestrially important ice feature—are loaded predominantly by wind, under global compressive stresses similar in magnitude to the shear stresses within glaciers (RichterMenge and Elder 1998). These bodies deform through a combination of creep and fracture, the latter process manifested by oriented leads or open cracks and by compressive shear faults which often form as conjugate sets that traverse a large fraction of the Arctic Basin (Kwok 1999; Schulson 2002). The cold, icy crust of Europa, an extraterrestrial feature which may shield an ocean beneath within which a form of life may exist or may once have existed (Reynolds et al. 1987; Hoppa et al. 1999; Pappalardo et al. 1999; Greenberg et al. 2000; Kargel et al. 2000), is loaded by the motion of diurnal tides and by non-synchronous rotation (Greenberg and Weidenschilling 1984). The crust deforms in a brittle manner, as evident from the networks of cracks that lace through it (Greeley et al. 2000). Of the two kinds of ice deformation, creep is probably the better known by readers of the geological and geophysical literature, and is certainly the more fully explored and better understood. The interested reader may wish to consult a number of excellent reviews of the subject (e.g., see Duval et al. 1983; Weertman 1983; Durham and Stern 2001). Fracture, in comparison, has only recently been systematically examined. The motivation with respect to terrestrial mechanics (Schulson 2001) comes largely from the growing recognition of the role it plays in limiting ice forces against off-shore engineered structures and in forming arctic leads (which serve as short-circuits paths for the transfer of heat from the ocean to the atmosphere during winter). With respect to Europa, the study of fracture is justified by the possible linking of the surface to the putative ocean. Yet, even though the field of ice fracture is relatively new compared to ice creep, much has been learned about how cracks nucleate and propagate within this material. Indeed, the study of ice has led to a new understanding of compressive failure of rock and other brittle materials (Renshaw and Schulson 2001), further justifying the effort. This paper considers the brittle failure of polycrystalline ice Ih, called ordinary or terrestrial ice1, at temperatures (>0.8 Tmp) and strain rates (~10-7 s-1 to 10-1 s-1) of engineering interest. The work emphasizes ice as a material and thus centers on structureproperty relationships. The objective is to elucidate the failure mechanisms. Tensile behavior is considered first, and is discussed in terms of the nucleation and the propagation of cracks. Brittle compressive failure, particularly under multi-axial loading, 1

Ice possesses 13 different crystal structures and two amorphous states. Ice Ih forms under terrestrial conditions and possesses the hexagonal crystal structure reflected in the shape of snow-flakes. The other crystallographic forms are thermodynamically stable only under high pressure.

1529-6466/00/0051-0008$05.00

202 

Schulson 

is addressed next, at greater length owing to its greater complexity. It is described in part by Coulomb's criterion and is discussed in terms of frictional crack-sliding and a new mechanism of shear faulting. The compressive ductile-to-brittle transition is then considered, and is explained in terms of the competition between crack-tip creep and crack propagation. The paper closes with a discussion of failure in the field and with evidence that the underlying physical processes may be independent of spatial scale. They may also operate in rock and other brittle materials. The work is based largely upon the research of the author and his students, with due regard for the literature. Before proceeding, it is worth noting that ice is unusual in exhibiting brittle behavior right up to the melting point, at deformation rates that are well below dynamic rates. Single crystals as well as polycrystals exhibit this behavior, implying that it originates in basic deformation processes and not through the restriction to slip imposed by grain boundaries. At root are sluggish dislocation kinetics (e.g., see Ahmad and Whitworth 1988; Shearwood and Whitworth 1991; Hu et al. 1995). Ice Ih slips preferentially on basal {0001} planes and this is impeded by the unique requirement of protonic rearrangement2 (Glen 1968). Molecular diffusivity is low as well—about three orders of magnitude lower than atomic diffusivity through metals at an equivalent homologous temperature3. This factor, however, appears to play more the role of suppressing diffusional creep within the coarsely grained material (~1 to 20 mm) that occurs naturally (Duval et al. 2000) than of impeding dislocation mobility. The “degree of brittleness” of polycrystalline ice Ih is rather high (Nixon and Schulson 1987), evident from the relatively low ratio4 of toughness (~1 J/m2) to surface energy (~0.1 J/m2). FAILURE UNDER TENSION Ice breaks in tension while bending under ships (Michel 1978) and above submarines (Bazant et al. 1995), while buckling to form pressure ridges (Hopkins 1998), while bending against the sloping sides of offshore structures (Riska and Tuhkuri 1995) and during the calving of icebergs (Hughes 1989; Rist et al. 1999). This section describes the failure characteristics and then discusses them in terms of deformation mechanisms. Characteristics Fresh-water ice. Equiaxed and randomly oriented aggregates of fresh-water ice (termed “granular ice”) fail under uniaxial tension via transgranular cleavage (Schulson et al. 1989a), at applied stresses around 1 MPa (Hawkes and Mellor 1972; Michel 1978; Currier and Schulson 1982; Schulson et al. 1984). The strength is essentially independent of strain rate, at least over the range 10-7 s-1 to 10-3 s-1, and is only slightly dependent upon temperature, increasing by dc) 2

(1)

Ice Ih is possibly unique in that translation of part of the crystal relative to the rest by the Burgers vector does not exactly reproduce the atomic arrangement of its protons. The translation introduces Bjerrum defects. The stress needed to create them (of formation energy 0.68 eV) is orders of magnitude greater than can be accounted for by the flow stress, and this suggests that some kind of protonic rearrangement probably occurs (Glen 1968). 3 The melting point diffusion coefficient of H2O in ice Ih is around 10-15 to 10-14 m2/s compared to 10-12 to 10-11 m2/s for elemental metals. 4 For structural ceramics and rock (see (Nixon and Schulson 1987), the ratio of toughness to surface energy is generally greater than 30.

 

Brittle Failure of Ice 

203 

Figure 1. Graphs of tensile fracture stress versus (grain size)-0.5 of equiaxed and randomly oriented polycrystals of ice Ih loaded under uniaxial tension (Schulson et al. 1984; Schulson 1987): (a) showing schematically the critical grain size, dc, below which the material exhibits a minor amount of inelastic deformation that increases with decreasing grain size; (b) showing that at –10°C at a strain rate of 10-3 s-1, where dc < 1.4 mm, the strength is controlled by crack nucleation; and (c) showing that at –10°C at a strain rate of 10-7 s-1,where dc > 6.7 mm, the strength is controlled by crack propagation.

or σt = Κd-0.5 (for d < dc) , (2) where σo, kt and Κ are material constants and dc is a critical grain size. Values (Schulson et al. 1984) at -10°C are σo = 0.6 MPa, kt = 0.02 Mpa√m and Κ = 0.044 Mpa√m; dc decreases with increasing strain rate, from >6.7 mm at 10-7 s-1 to 1.6 mm at 10-6 s-1 to (3 π γ G/8[1 – ν])0.5

(4)

where γ is the surface energy (109 mJ/m2, (Ketcham and Hobbs 1969), G is the shear modulus (initially taken to be 2300 MPa; a better value is 3500 MPa, (Gammon et al. 1983), n is Poisson's ratio (0.33, (Gammon et al. 1983) and <m> is the Taylor orientation factor5. The problem is that the better values of G and <m> yield kt ≈ 0.14 Mpa√m, which is about a factor of seven higher than the measured value of 0.02 Mpa√m. The current view, therefore, is that the pile-up mechanism is not the important one, at least at the higher rates of deformation where crack nucleation governs the tensile strength. Instead, grain boundary sliding probably dominates. Direct observations, in fact, have been made of crack nucleation through grain boundary sliding, albeit under compression (Nickolayev and Schulson 1993; Picu and Gupta 1995a,b; Weiss and Schulson 2000), underlining its importance.   Interestingly, the tensile strength of granular ice, when governed by crack nucleation, can double upon tension-compression cycling (Cole 1990) to stresses significantly above σ0 (Δσ ≈ ±1 MPa). The cycling evidently lessens the effectiveness of localized stress concentrators, leading to higher applied stresses for crack nucleation. Ledges on grain boundaries, for instance, have been directly observed to emit dislocations (Liu and Baker 1995; Baker 1997) under monotonic loading. Presumably under cyclic loading, such sites 5

In earlier work (Schulson et al. 1984); (Schulson 1987) it was assumed that <m> ≈ 3, the value commonly used for f.c.c. metals which exhibit a high degree of slip multiplicity. A higher value (~5) is more appropriate for hexagonal crystals which possess only two independent slip systems.

206 

Schulson 

also allow strain to be accommodated, lessening the intensification of stress there. One wonders whether alumina and other ceramics might be strengthened through similar pretreatment. Crack propagation. Cracks within “more finely grained” (d < dc) ice do not propagate immediately upon nucleation. Instead, the applied stress must increase until the mode-I stress intensity factor reaches the critical level. During the process, more inelastic deformation occurs and so, more cracks nucleate. Finely grained aggregrates (d ~ σ33 (Fig. 5a), terminal failure occurs by irregular splitting in the loading plane; i.e., on a plane parallel to the free surface. Along these paths, the along-column confining stress has no systematic effect on the strength, taken as the across-column stress at

Figure 5. Schematic sketches showing four paths of proportionally loading S2 ice under biaxial compressive stresses; plus the failure mode: (a) R21=0, σ11>σ33; (b) R21=0, σ33>σ11; (c) R31=0, R21R21c.

 

Brittle Failure of Ice 

213 

terminal failure, σ11f (Smith and Schulson 1994; Schulson and Nickolayev 1995; Schulson and Gratz 1999). Similarly, when σ33 > σ11 (Fig. 5b), failure again occurs by splitting in the loading plane and the confining stress (this time σ11) again has no effect on the strength (σ33f). The strength, however, is greater along the columns than across them, by about a factor of two (Kuehn and Schulson 1994b). Along both kinds of path, the confining stress has little or no component resolved onto the macroscopic failure plane, accounting for its nil effect on the failure stress. Table 1. Characteristics of columnar ice. Characteristic

Fresh-water ice

Salt-water ice

Grain shape

columnar

columnar

Texture

S2

S2

Column diameter (mm)

7.6±1.6

8.5±1.8

Salinity (%)

—

4.5±0.3

Density (kg/m3)

917.7±0.4

910±3

Porosity (%)

~0.1

3.9

Brine pocket spacing (mm)

—

0.6±0.2

Pore size (mm)

—

~0.5

When S2 ice is biaxially loaded under two across-column stresses, the confining stress (denoted σ22) has a large effect. Two cases are distinguished, depending upon the ratio of the minor (i.e., confining) stress to the major stress, R21 = σ22/σ11, relative to a critical degree of confinement, R21c: (i) Strengthening: For lower confinements R21 < R21c (Fig. 5c), the strength, taken as σ11f, increases sharply with increasing confinement. Figure 6 (branch AB) illustrates this effect. It can be described by the relationship: σ11f = σu + k σ22 = σu/(1 – k R21)

(6)

where σu is the uniaxial across-column strength and k is a dimensionless constant; their values increase with decreasing temperature and are similar for both types of ice, Table 2. Terminal failure occurs by shear faulting (R21 > 0) across a plane parallel to the columns, but inclined (at -10°C) by 28±4° to the major stress (more below). That the fault does not lie in a plane parallel to the intermediate principal stress (i.e., σ22) does not contradict a basic tenet (Rudnicki and Rice 1975) of faulting, because S2 ice is an anisotropic material. Within this lower confinement regime, the materials deform inelastically in essentially a plane strain manner, with little or no displace-ment in a direction parallel to the long axis of the columns.   (ii) Weakening: Under higher confinements R21 > R21c (Fig. 5d), the strength decreases with increasing confining stress; branch BB, Figure 6. This effect can be described by the relationship: σ11f = σ0 – σ22 = σ0/(1 + R21)

(7)

where σ0 is another constant, Table 2, which again is similar for the two types of ice. Temperature is still a factor, but a less important one. Failure now occurs by

214 

Schulson  Figure 6. Brittle compressive failure envelopes for S2 fresh-water ice of ~6 mm column diameter biaxially loaded acrosscolumn at -10°C and -40°C at 10-2 s-1 (major strain rate) showing strengthening with increasing confining stress under lower degrees of confinement (branch AB, A'B') and weakening with increasing confining stress under higher degrees of confinement (branch BB, B'B'). The envelope is symmetric with respect to the principal stresses because S2 ice is isotropic in the X1-X2 plane. Data are thus plotted twice about the line σ11=σ22 (From (Smith and Schulson 1993).

cleavage in and then spalling out of the loading plane. Cracks propagate across the columns, through a mode analagous to that observed during higher-speed, throughthickness, across-column indentation of S2 ice where lateral confinement is provided by material surrounding the contact zone (Grape and Schulson 1992; Sodhi 2001). Note that R21c decreases with decreasing temperature, Table 2. Table 2. Experimental constants for Equations (6) and (7) describing the biaxial compressive strength of fresh-water and salt-water S2 ice at a maximum principal strain rate of 10-2 s-1 (Smith and Schulson 1993, 1994; Schulson and Nickolayev 1995; Pradeau and Schulson 1998). Ice fresh-water ice

salt-water ice

T (°C)

σu (MPa)

k

σ0 (MPa)

R21c

- 10

3.8±1

3.1

15

.25

- 40

10±1

~4

20

~0.1

- 2.5

2.7±0.8

1.8

11

0.45

- 10

3.5±0.7

3.1

16

.25

- 40

6.5±1

4.8

16

~0.1

Triaxial compressive strength: Confinement under triaxial loading raises the compressive strength. Three regimes are evident, depending upon the loading path (Gratz and Schulson 1997; Schulson and Gratz 1999); see Figure 7. (i)

Regime 1: When loaded mainly across the columns under a small across-column confining stress (R21 ≤ 0.25, Fig. 7a), the strength, σ11f, is independent of the degree of confinement along the columns (Fig. 8a) but increases sharply with increasing across-column confinement (Fig. 8b). Failure occurs by macroscopic shear faulting on a plane parallel to X3, but inclined (at -10°C) by 31±6° to X1. The behavior here is essentially identical to the behavior under lower-confinement, across-column, biaxial loading. Again, the along-column component of the confinement has little or no component resolved onto the failure plane and thus does not affect the strength.

 

Brittle Failure of Ice 

215 

Figure 7. Schematic sketches showing three regimes of brittle compressive failure of S2 columnar ice proportionally loaded triaxially across and along the columns: (a) Regime-1; (b) Regime-2; (c) Regime-3.

(ii)

Regime 2: When loaded mainly across the columns, but under higher degrees of across-column confining stress (R21 > 0.25; Fig. 7b), the strength increases markedly upon increasing the along-column confining stress (Fig. 8c). Failure occurs by faulting on planes that cut the columns (at -10°C) at 33±13° to their long axis. Under higher confinements (R31 > 0.1, termed Regime -2b, Fig. 8c), the strength becomes less dependent upon confining stress, exhibiting more the character of ductile failure than of brittle failure. Correspondingly, the fault is more steeply inclined, cutting the columns (at -10°C) at 41±3°. The reduction in the sensitivity to confinement corresponds to the hydrostatic component of the applied stress (σ11 + σ22 + σ33)/3 reaching ~14 MPa and appears to mark the onset of dynamic recrystallization (Schulson and Gratz 1999), at least at -10°C where experiments have been performed to date.

(iii) Regime 3: When loaded mainly along the columns (Fig. 7c), the strength, σ33f, increases in proportion to the smaller of the two across-column confining stresses (Fig. 8d). Failure now occurs by faulting on a plane inclined (at -10°C) 22±9° to X3 and parallel to X1. Within each regime, the strength, sf, can be described by a relationship of the form: σf = σu' + k'σc

(8)

where σu' and k' are constants and σc is the appropriate confining stress. Table 3 lists their values. Note that within Regime 1 the constant k' of Equation (8) has the same value as the constant k of Equation (6). Note also, as already mentioned, that the brittle compressive strength of the salt-water ice is essentially indistinguishable from that of fresh-water ice. The only detectable difference (Schulson and Gratz 1999) is that the shear fault is narrower in the salt-water material—i.e., less than one grain diameter vs. two to three diameters—owing to a crack-stopping effect of its pores.

216 

Schulson 

Figure 8. Graph of brittle triaxial compressive strength of S2 fresh-water and salt-water (4.5 ppt salinity) ice at -10°C at 6×10-3s-1 (maximum principal strain rate): (a) versus along-column confinement under low across-column confinement at three different levels, Regime-1; (b) versus across-column confining stress for various levels (01 m s-1, creating a band of fragmented ice or “gouge” and ending in terminal failure. Figure 17 sketches the process as it is thought to occur on the small scale. [Figure 25, below, shows evidence of its occurrence on the geophysical scale.] The stress to trigger the fault is assumed to be the applied stress that causes a nearsurface microcolumn to break. Accordingly, a first-order analysis (Schulson et al. 1999) suggests that each “tooth” in the “comb” is subjected to both an axial force P per unit depth and a moment M per unit depth, Figure 18. The moment is the more important factor. It is caused by the shear stress σs,eff effective in causing sliding, and is equated to the difference between the applied shear stress σs resolved onto the sliding plane and the frictional resistance μσn induced by the stress σn normal to the sliding plane. The cracks that define the fixed end of the microcolumns experience mixed-mode (I and II) loading and propagate when the mixed mode stress intensity factor reaches a critical level. Assuming that the situation is similar to the case of an edge crack in a brittle plate and that KIic = KIc for ice (Shen and Lin 1986) , then from the analysis of Thouless et al. 1987 it can be shown that the lower limit (under uniaxial loading) of the terminal failure stress may be approximated by the relationship (Schulson et al. 1999): 2KIc σf = (17) 0.5 (3βh ) (1− μ) where β = h/w is the slenderness ratio of the microcolumns. Under confined loading, following a derivative (Schulson et al. 1991b) of the Ashby and Hallam (1986) analysis, the failure stress may be roughly approximated by11: 2K Ic σf = (18) 0.5 (3βh ) {(1 − μ ) (1 − R)− 2R μ} where R is the appropriate confining ratio and where the fuller term in the denominator expresses the impediment to parent crack sliding introduced by the confinement. By 11

The denominator in Equation (18) tends to zero when μ tends to (1 – R)/(1 + R). This is not a major weakness because the model only applies to cases of low confinement. Lehner and Kachanov (1996) offer an analysis which overcomes this shortcoming.

 

Brittle Failure of Ice 

231 

Figure 17. Schematic sketch of the initiation and growth of a macroscopic compressive shear fault, from the onset of terminal failure (i.e., from very near the peak of the stress-strain curve). The lighter, hexagonal background denotes grain boundaries. (a) Before initiation, damage is uniformly distributed throughout the matrix, and consists of a mixture of parent grain boundary cracks inclined by about 45° to the highest principal stress, parent cracks with wing-crack extensions, and sets of secondary "splay cracks" stemming from one side of parent cracks. The parent, inclined cracks plus the "splay" constitute a "comb-crack". (b) Near-surface "comb-crack" fails, owing to less constraint there, shedding load and redistributing stress to adjacent material within which new cracks initiate. The nucleus of the eventual shear fault is formed. (c) More comb-cracks fail and localized damage spreads as a band across the section, creating granulated material or gouge in its wake and forming new damage in a "process zone" ahead of the advancing fault front. (d) A recognizable macroscopic shear fault (from Schulson et al. 1999).

232 

Schulson 

Figure 18. Schematic sketch of the "comb-crack" fault initiation mechanism. Secondary cracks initiate from one side of parent cracks which are inclined by angle ψ to the maximum (most compressive) principal stress. The secondary cracks create sets of slender micro-columns of length h and width w, fixed on one end and free on the other. Axial loading produces a longitudinal load/unit depth P on the microcolumns, and frictional sliding across the parent, inclined cracks under effective shear stress σs produces a moment/unit depth M on the microcolumns. M is more important than P and eventually becomes large enough to break the microcolumns, preferentially near a free surface, thereby triggering the growth of a macroscopic shear fault across the load-bearing section. From Schulson et al. (1999).

setting12 KIc = 0.1 MPa m , μ = 0.5, β = 5.3±2 (length and width, respectively, in the case illustrated above (Fig. 16), are 6.2±4.0 mm and 1.2±0.2 mm) and h = d/2 = 5 mm (above observations) and R = 0, one finds that σf = 1.5±0.5 MPa. This estimate is of the correct order of magnitude, but is about a factor of three too small. The discrepancy reflects perhaps the exclusion of crack interactions and of additional friction that arises through the rubbing together of adjacent microcolumns within the comb. The model captures the general character of the brittle compressive strength. It embodies (grain size)-0.5 dependence through the dependence of the column length on grain size. It captures strain-rate softening through the friction coefficient, which, as already noted, decreases with increasing sliding speed (Kennedy et al. 2000): for instance, an increase in velocity from 10-4 to 10-2 m/s , which corresponds roughly to an increase in strain rate from 10-3 to 10-1s-1, lowers μ from 0.5 to 0.2 (at -10°C), leading to an expected reduction in strength of ~40%, in fair agreement with observation (Fig. 2). The model accounts for thermal softening, also in terms of the friction coefficient. For instance, the reduction from μ = 0.8 at -40°C to μ = 0.5 at -10°C leads to an expected reduction in strength of fresh-water ice of about 50%, in fair agreement with observation (Figs. 3 and 6). The friction coefficient (at least under low normal stress) of saline ice at -40°C is ≈ 15% lower than that of fresh-water ice (Kennedy et al. 2000), accounting in part for the lower strength of saline ice at the lower temperature. The model also dictates the ratio 1/(1 – 3R) for the confined strength to the unconfined strength of S2 ice at -10°C under lower across-column confinement (R21 < R21c), in good agreement with measurement and with Coulomb’s criterion [see section titled Failure criteria, above]. 12

The appropriate value of fracture toughness may be smaller than that assumed, owing to size effects. The lowest possible value is set by surface energy through the relationship Kc = [2 γ E/(1 – ν2)]0.5. This gives Kc ~ 0.05 MPa m . Whichever value is used does not change the sense of the argument. 

 

Brittle Failure of Ice 

233 

Lastly, it dictates that the critical confinement ratio R21c = (1 – μ)/(1 + μ) = 0.55, 0.33 and 0.14 at -2.5°C, -10°C and -40°C, respectively, in reasonable agreement with observation [see section titled S2 columnar ice, above]. Although more work is needed, the combcrack model appears to capture the essence of the faulting process. It is interesting to note that recent modeling of faulting in rock (Gupta and Bergström 1998) invoked both a Weibull-like distribution of failure stresses and a stress enhancement factor resulting from crack interactions. In so doing, the work simulated rather closely the evolution of the process, including the “process zone” observed in granite (Lockner et al. 1991). What it did not elucidate, however, was the nature of the “trigger.” Like earlier models, the work was founded upon observations (by other investigators) that were made using indirect techniques (e.g., acoustical emission, computed X-ray tomography) which have relatively coarse resolution, exacerbated by complex microstructures. An elliptical nucleus within which the shear modulus was reduced through localized cracking was thus assumed. Also, the model did not capture initial blowout at the free surface (Lockner et al. 1991), and, as its authors note, contains several free parameters. In comparison, the “comb-crack” mechanism incorporates a micromechanical trigger that is based upon direct, experimental observation, it accounts for blowout, and it has only one free parameter (the slenderness ratio of the microplates). How are comb-cracks created? Contact, it appears, is essential across a fractured interface. This is evident from the observation that secondary cracks generally do not form on pre-engineered parent cracks, created using Teflon tape (Schulson et al. 1999). This could explain their absence in earlier modeling of brittle compressive failure, where sliding “cracks” were created in glass using slots (Brace and Bombolakis 1963) and in Columbia resin using metallic plates lined with Teflon (Horii and Nemat-Nasser 1985;1986). One possible process is Hertzian cracking, arising from stresses that develop at randomly arrayed asperities along cracked grain boundaries. The one-sided characteristic could then be explained in terms of different tensile strengths on either side of the boundary, owing to different crystallographic orientations. This kind of process is considered to be unlikely, however: the spacing of the secondary cracks seems to be too regular to have arisen from randomly arrayed points of contact; and secondary cracks are not limited to cracked grain boundaries: they also form transgranularly within S2 columnar ice (see Fig. 11; Schulson and Gratz 1999). Moreover, should their origin be similar to the origin of splay cracks in rock, the one-sided characteristic could not be explained in terms of small-scale crystallography. A more likely explanation involves non-uniform sliding. The idea here, extending earlier suggestions (Martel and Pollard 1989;Cooke 1997) in connection with the fracture of rock, is that a gradient of displacement along the parent crack, ∂u1/∂x1 , creates normal stresses, of magnitude proportional to the gradient, that act across planes perpendicular to both the sliding plane and the sliding direction. Compressive stresses then develop on one side, while tensile stresses develop on the other. Figure 19 illustrates this notion. There, the uphill material that borders the upper side of the parent crack is displaced more than the downhill material, and so the upper-side material experiences a compressive stress while the underside material experiences tension. Should non-monotonic displacements develop, the tensile stress field would alternate to the other side of the sliding crack. Barring creep relaxation (see section below entitled Ductile-to-Brittle Transition Under Compression), the tension, once great enough, is relieved through the formation of secondary cracks. (Note that the secondary cracks in Figure 16 above and in Figure 26 below are not perpendicular to the parent, sliding crack.) The applied stress under which forms the set that constitutes a comb-crack appears to be approximately constant, given that one set formed within 10-3 s (i.e., in the interval between adjacent images on a high-

234 

Schulson 

speed film (Smith and Schulson 1993). The secondary cracks are imagined to form sequentially, starting near one tip of the parent crack and releasing proportionately more strain energy, proportional to K112 and thus to the distance from the tip of the parent crack to the initiation site of the secondary crack. Their length is thus imagined to increase with distance from the tip. Comb-crack A of Figure 15, for instance, could have formed in this manner.

Figure 19. Schematic sketch showing the proposed sequential (a-e) creation of secondary cracks under compression through non-uniform sliding across an inclined, cracked grain boundary. The maximum (most compressive) principal stress acts along the vertical direction.

Regardless of their origin, the important point is recognition of a new frictional sliding mechanism and of the role it appears to play in destabilizing damaged material. This constitutes a new element in the understanding of brittle compressive failure. Although derived from a study of ice, the comb-crack mechanism is not limited to that material. DUCTILE-TO-BRITTLE TRANSITION UNDER COMPRESSION

The ductile-brittle transition (Fig. 2) marks the point where the compressive strength reaches a maximum (with respect to strain rate). It thus limits the loads exerted by floating ice features against engineered and other structures. The objective now is to consider its origin. Definition

The transition is defined macroscopically, in terms of the shape of the stress-strain curve and the behavior/appearance of the material. Accordingly, ice is considered to be ductile when its stress-strain curve exhibits either a rounded peak or a plateau and the material does not suddenly lose its load-bearing ability, even though it may become so highly damaged that initially transparent material becomes milky-white in appearance. On the other hand, ice is considered to be brittle when its stress-strain curve is pseudolinear in shape and terminates suddenly owing to the onset of a mechanical instability.

 

Brittle Failure of Ice 

235 

More fundamental definitions of the transition are possible, involving, for instance, the competition between the emission of dislocations from near crack tips and crack extension. However, for the present purpose, the macroscopic definition suffices. The aim is to model the strain rate at which the transition occurs. Generally, this happens over a range of strain rates (about a factor of three). For the purposes of this discussion, however, the transition is defined in terms of a specific strain rate ε&t . As noted above, this is typically 10-4 to 10-3 s-1 (in small test specimens), but increases with decreasing grain size, increasing salinity, increasing confinement and increasing pre-deformation.

Figure 20. Schematic sketches of log strength (σ)- log strain rate ( ε& ) showing the effects of grain size, salinity, confinement and compressive pre-strain on the ductilebrittle transition strain rate, ε&t .

Transition models

Two approaches have been developed (Schulson 1990): one is phenomenological and is based upon the idea that the transition occurs at the strain rate, ε t , where the ductile strength equals the brittle strength, σd = σb (Fig. 20). As noted above, the ductile failure stress is characterized by strain-rate hardening and obeys Hill’s (1950) criterion for the yielding of plastically orthotropic materials. The brittle strength, as already

236 

Schulson 

discussed, decreases with increasing strain rate. [The argument can be made as well by assuming that the brittle strength is rate independent.] The variation of ε t with the factors noted in Figure 20 can then be understood in terms of their effects (noted above in Overview in the section entitled Brittle Failure Under Compression) on the respective strengths. Salinity/porosity (Buck et al. 1994) and pre-deformation (Kuehn et al. 1988) lower the ductile strength, but not the brittle strength, thereby increasing the transition strain rate. Grain refinement increases the brittle strength (Schulson 1990), but not the ductile strength (Schulson and Cannon 1984; Cole 1985), again raising the transition rate (Batto and Schulson 1993). Confinement increases the brittle strength more than the ductile strength (Schulson and Nickolayev 1995) and so raises the transition rate still further. These effects have been modeled (Schulson and Buck 1995; Schulson and Nickolayev 1995) and reasonable agreement with observation has been obtained. The other approach considers micromechanical processes. An early view was that the transition occurs when the ice first begins to crack, as is the case under tension. Under compression, however, ice begins to crack, at strain rates two to three orders of magnitude below the macroscopic transition strain rate, and so crack nucleation cannot account for the transition. Another possibility is that the transition coincides, in some way, with a certain proportion of intergranular cracks. This is based upon the observation (Hallam et al. 1987; Kalifa et al. 1989; Gold 1997) that the ratio of the number density of intergranular to the transgranular cracks increases with increasing strain rate within the vicinity of the transition. The problem with this notion is that the strain rate marking any given ratio of grain-boundary cracks to total cracks increases with increasing grain size (see Fig. 2 of Gold 1997), implying that the transition strain rate should also increase with increasing grain size. The effect of grain size is just the opposite (Batto and Schulson 1993). Schulson's view (Schulson 1990) is that the transition marks the point where cracks begin to propagate, albeit in a stable manner. In this scenario, a competition is imagined between stress relaxation and stress build-up at crack tips. Relaxation through crack-tip creep is assumed to dominate at lower deformation rates, inhibiting propagation. At higher rates stress-buildup dominates, allowing the mode-I stress intensity factor to reach the critical point. Indeed, experiments (Batto and Schulson 1993) on specimens containing inclined “Teflon cracks” from whose tips wing cracks were initiated show that cracks do not propagate at strain rates below the transition, but do propagate at rates above the transition rate. Similarly, comb cracks do not develop on the ductile side of the transition (Iliescu and Schulson, unpublished), implying that the necessary tensile stressses relax. These observations support Schulson's suggestion, and so it is the one pursued. Imagine a crack-tip creep zone of radius, r, defined as the region within which creep strain exceeds elastic strain (Fig. 13). Although brittle failure is attributed more to comb cracking than to the propagation of wing cracks, the analysis is made for wing cracks because they are more easily analyzed. It is assumed that similar stress states and creep zones characterize the near-tip region of the two kinds of crack. The model incorporates the hypothesis that cracks propagate when r < 2 fa, where 2a is the length of the parent, sliding crack and f < unity. The criterion for the transition may then be expressed as: r = 2 fa . (19) The creep zone radius can be obtained from the Riedel and Rice (1980) model of creep zones around non-interacting cracks loaded under far-field tension by invoking secondary or power-law creep ( ε& = Bσ n , where n = 3). Thus: 2 n K 2I ⎛ ( n + 1) E Bt ⎞ ⎜ ⎟ r= 2πE 2 ⎜⎝ 2nα( n +1) ⎟⎠

2 ( n −1)

F

(20)

 

Brittle Failure of Ice 

237 

where F and α are of order unity; t is the loading time, and is approximated as the ratio of & , where the loading rate, the critical stress intensity factor and the loading rate, t ≈ K c K I & & K I , is obtained through the product of partial derivatives K = ( ∂K ∂σ )( ∂σ ∂ε )( ∂ε ∂t ) . The derivative ∂σ/∂ε is essentially a modified Young's modulus E' (i.e., E reduced by damage) and ∂ε /∂t is the applied strain rate. The other derivative depends on the details of the crack geometry and here is taken from wing crack mechanics (Ashby and Hallam 1986) to be of the form (Weiss and Schulson 1995) ∂K ∂σ= ( πa 3)

0.5

(

⎡ 1 − R 1+μ 2 ) ⎢⎣(

)

0.5

− μ (1+R ) ⎤⎥ ⎦

where 2a = d. Upon combining these relationships, upon replacing KI in Equation (20) with KIc on the grounds that the secondary cracks are injected under load and are on the verge of growth, and upon assuming E' ≈ E, the transition strain rate may then be expressed as (Schulson and Nickolayev 1995): ε& t =

BK 3Ic 1.5

fd

(

⎧ ⎡ 2 0.5 2 ⎨ ⎡⎣1 + μ ⎤⎦ − μ − R ⎢μ + 1 + μ ⎣ ⎩

)

0.5 ⎤ ⎫

(21)

⎥⎦ ⎬⎭

for R < (1 – μ)/(1 + μ). For larger confinement a different treatment is required owing to the change in failure mode. The transition is thus expressed in terms of independently measurable materials parameters (B, Kc, μ and d) and in terms of a fraction, f, whose value was calculated (Batto and Schulson 1993) to be f ≅ 0.015, from the Riedel-Rice (Riedel and Rice 1980) model of crack-tip creep and from separate measurements. Schulson's model appears to capture all of the observations to date: the order of magnitude of the transition strain rate, as well as its measured dependence (Batto and Schulson 1993) on (grain size)-1.5±0.5 (Fig. 21); the increase in the transition strain rate with increasing confining stress under lower confinement R21 < R21c (Schulson and Buck 1995; Schulson and Nickolayev 1995), as well as the order-of-magnitude higher transition strain rate of salt-water ice versus fresh-water ice13 (Fig. 22). Albeit qualitatively, the model also captures the order-of-magnitude increase in the transition strain rate induced by compressive pre-deformation (Kuehn et al. 1988): the pre-strain produces damage and this raises the creep constant (Weiss 1999) without changing the exponent n. Finally, when cognizance is taken not only of the decrease in the creep, B, constant with decreasing temperature, but also of the large increase in friction coefficient (Kennedy et al. 2000) and of the small increase in fracture toughness (Nixon and Schulson 1988; Dempsey 1996), the model accounts for the order of magnitude reduction (Qi and Schulson 1998) in the transition strain rate upon reducing temperature from -10°C to -40°C. Equation (21) also accounts for an observation not previously explained. In their early work on compressive failure, Hawkes and Mellor (1972) did not report a sudden ductile-to-brittle transition, at strain rates up to 10-2 s-1 at -7°C. The reason for its absence, it now appears, is that the grain size of their material (~0.7 mm) was so fine that it raised the transition to a strain rate of ~0.7×10-2 s-1; i.e., to very near the highest rate they explored. Given the limited data and given the fact that the transition occurs over a

13

The higher transition rate for salt-water ice stems from a higher creep rate, reflected in the creep constant: B = 5.1×10-6 MPa-3 s-1 for salt-water ice of 4.5 ppt salinity and B = 4.3×10-7 MPa-3 s-1 at -10°C (Sanderson 1988). The fracture toughness (Dempsey et al. 1999) and the coefficient of friction (Kennedy et al. 2000) are essentially the same for fresh-water and salt-water ice.

238 

Schulson  Figure 21. Graph showing the measured and the calculated effect of grain size, d, (i.e., column diameter) on the ductile-brittle transition strain rate at –10°C of fresh-water S2 ice loaded uniaxially across the columns. Calculations were made using Equation (20) and the parametric values B = 4.3×10-7 MPa-3 s-1; KIc = 0.1 MPa m ; μ = 0.5; f = 0.015. The dotted lines are drawn through the data and show possible d-1 and d-2 functionalities. The solid line is the one calculated from the model. From Batto and Schulson (1993).

Figure 22. Graphs showing the measured and the calculated effect of across-column confining stress on the ductile-brittle transition strain rate at -10°C of both fresh-water and salt-water (of 4.5 ppt salinity) S2 ice of ~6 mm column diameter loaded biaxially across-column. The calculations (curve) were made using Equation (20) and are limited to the regime of lower confinement where failure occur via shear faulting. Parametric values: B = 4.3×10-7 (fresh-water ice) and 5.1×10-6 (salt-water ice) MPa-3 s-1 (Sanderson 1998); KIc = 0.1 MPa m (both materials); μ = 0.5 (both materials); d = 6 mm (both materials); f = 0.015 (both materials) (from Schulson and Buck 1995; Schulson and Nickolayev 1995).

range of strain rates, it seems likely that experiments approached, but did not fully enter, the brittle regime. Indeed, they (Hawkes and Mellor 1972) noted that the compressive strength appeared to reach a “limiting value” at 10-2 where the specimens failed by splitting. The agreement between theory and experiment is probably better than it should be, given the assumptions that are invoked. The Riedel-Rice analysis was developed for creep at the tips of non-interacting cracks within an isotropic, homogeneous material

 

Brittle Failure of Ice 

239 

loaded under far-field tension. It may not be strictly applicable to crack-tip creep within tensile zones induced locally within a damaged, plastically anisotropic aggregate loaded under far-field compression. Also, stress relaxation within the crack-tip region probably occurs through a combination of primary and secondary creep, and not just secondary creep as assumed. There is also the question of friction and whether coefficients obtained from ice sliding upon ice, smoothly surfaced, are similar to those for sliding cracks. Nevertheless, in incorporating resistance to creep, to fracture and to frictional sliding, as well as the size of the stress concentrator, the model appears to capture the essential physical and microstructural elements underlying the ductile-to-brittle transition. ON THE FORMATION OF “LEADS” IN THE ARCTIC SEA ICE COVER

Leads, or open cracks, are important features of sea ice covers at high latitudes. They serve as passageways for the movement of ships and submarines, and they act as lowresistance paths for the transfer of heat from the ocean to the atmosphere. Some leads are so long that they span a large fraction of the Arctic ocean (Kwok et al. 1992, 1999). How they form and the role they play in the subsequent deformation of the cover are issues that relate directly to polar marine transits and to global climate (Maykut 1982; Johanessen et al. 1995). The objective now is to consider their origin. Schulson and Hibler (1991) first suggested that scale-independent compressional failure processes may be at play, given wing-like cracks in the ice cover on the Beaufort Sea (Fig. 23). Although controversial, that is the path followed here.

Figure 23. Satellite image of part of the sea ice cover on the Beaufort Sea off the coast of Alaska (Banks Island is just to the right of the image) 11 February 1983. The dark smudge at the top of the image is from cloud-cover. Winglike cracks, stemming from the tips of inclined leads, are arrowed (from Schulson and Hibler 1991). Compare with wing-cracks generated in the laboratory, Figures 12 and 14.

Satellite images of high-latitude sea ice covers reveal patterns of cracks, often aligned and generally characterized by intersecting features termed “slip lines” (Marko and Thomson 1977; Erlingsson 1988; Kwok et al. 1992; Lindsay and Rothrock 1995; Overland et al. 1998) (Fig. 24a). [The figure caption describes the proposed order of shearing.] The “lines” are actually narrow (~0.5 km to 5 km) bands of damage ~50 km apart (Overland et al. 1998). They intersect over a range of angles, commonly focused around 40° (included angle), thereby forming diamond-shaped patterns. Since “slip lines” in the covers do not form orthogonal sets, they are different in character from the slip lines of metal plasticity theory (see e.g., Hill 1950). However, they do delineate diffuse boundaries diffuse boundaries across which discontinuous displacements occur, as evident from analysis of synthetic aperture radar images (Overland et al. 1998; Kwok et al. 1999). The “slip lines,” it is suggested, are the geophysical analogue of the compressive shear faults that form within plates of S2 ice biaxially loaded across-column

240 

Schulson 

Figure 24. Photographs showing sets of intersecting faults on scales large and small: (a) Satellite image of the sea ice cover in the Beaufort sea off the coast of Alaska 25 September 1974. From Marko and Thomson (1977). Fault-1 probably formed before fault-2, and fault-3 before fault-4. Displacements of fault-1 at X probably occurred by shearing along fault-2, and displacement of fault-3 at Y by shearing along fault-4.

to terminal failure under moderate across-column confinement (Fig. 24b, next page). Covers of first-year arctic sea ice, it is recalled, are largely made from S2 ice and are stabilized against buckling by a potential energy gradient. Slip lines appear to develop through a process of progressive damage to the parent ice cover, analogous to the development of shear faults in the laboratory. The development within the pack ice is expected to be more complicated, however. There the parent sliding cracks are imagined to be narrow, refrozen leads from an earlier time that failed. A recent analysis (Hibler and Schulson 2000) considered the cover to be an anisotropic continuum composed of thin and thick ice. Within each element (~10 km on edge) are located narrow, refrozen leads of random orientation. These refrozen leads

 

Brittle Failure of Ice 

241 

Figure 24. (b) Shear faults in S2 fresh-water ice of ~3 mm column diameter compressed biaxially across-column (R21 ≈ 0.1) in the directions shown, at -10°C at 10-3 s-1. View is along the columns.

contain thin ice and are embedded within surrounding thick pack ice. The thin, “old leads” are imagined to “fail” when the local biaxial compressive stresses satisfy the appropriate failure criterion. The criterion taken was based upon a compressive failure envelope that was generated in the laboratory (Schulson and Nickolayev 1995), similar to the ones shown in Figures 6 and 10. From a random array of such pre-existing flaws, the ones that fail preferentially are those of a particular orientation with respect to the maximum principal stress of the far-field applied stress. The critical orientation depends upon both the ratio of the minor to the major compressive stress and friction. Upon failing, the flaws redistribute stress in their vicinity, leading to fresh failures in adjacent elements. In this way is created a kind of macroscopic “process zone” within which minileads alternate in orientation as they form ahead of the advancing front. The mini-leads eventually link up, in a manner not yet specified but which probably depends upon the details of the local failure process, as it does on the smaller scale. Once linked, a band of damage extends all the way across the cover, thereby creating a slip line, itself oriented with respect to the maximum principal stress. The slip line, in other words, appears to be created through a process of consecutive cracking, and the damage band, which eventually defines the fault, propagates across the cover. Figure 25 shows satellite images which appear to support this proposed process. The propagation velocity is −˜ 13 km / day , which curiously is within an order of magnitude of the shear fault velocity in the laboratory (see section Shear faults and comb-cracks, above). Slip lines divide the parent cover into individual floes whose size is set by the spacing between them. When the failure criterion of the gouge between them is satisfied, the floes slide with respect to each other, deforming little or not at all (Overland et al. 1998; Kwok et al. 1999), except perhaps by grinding away at the interface and creating additional gouge material. Slip continues, it is imagined, until either the wind stops or

242 

Schulson 

(a)

(b)

Figure 25. Synthetic aperture radar images (ERS-1) of the ice cover in the Beaufort Sea, taken by satellite (a) on day 79 (at 22:35) 1992 and (b) on day 83 (at 6:51) 1992. Note the "damage zone" ~5 km wide that started at the lower, left-hand corner and propagated ~40 km in the 3-day interval. Localization of deformation occurs within this zone. Outside the zone, deformation appears to have ceased, judging from the negligible change in the density of the distributed cracks over the 3-day interval. The damage zone defines the beginning of a shear fault, evident from the relative shear displacements (arrows) that occur across the fault. [Cell size = 5 km.]

 

Brittle Failure of Ice 

243 

changes direction, or until healing impedes sliding to the point that it “freezes” the process. At that stage, a new fault is expected to form, possibly of an orientation different from the first. Once formed, sliding again occurs when the failure criterion for the gouge material of that fault is satisfied. The process then repeats itself. It is thus suggested that the ice cover exhibits a mixture two kinds of inelastic deformation (Fig. 26). One pertains to a continuum (type-I), albeit damaged, and the other (type-II), to a granular medium. Deformation of the continuum ends in terminal failure, of the kind described above. Deformation of the granular medium leads to stick/slip behavior (noted above, in the Overview of the section Brittle Failure Under Compression, for the small scale), manifested by the kind of wildly oscillating pack-ice stresses recently reported (Richter-Menge and Elder 1998). Both kinds of deformation are frictional in character, but the nature of the “friction coefficient” is probably different. The interplay between the two is thought to govern the behavior of the sheet. Under what level of stress might the field cracks have formed? That depends upon the mechanism. The similarity in the appearance of some leads to small-scale wing cracks

Figure 26. Schematic sketch of two types of deformation: type-I depicts ice as a continuum; type-II, as a granular material. A fault/lead develops during terminal failure (b,c), creating gouge (c) across which frictional sliding occurs at lower applied stresses. Eventually, the fault/lead heals or the wind changes direction (d) and the ice again behaves as a continuum.

244 

Schulson 

(a)

(b)

(c) Figure 27. Photographs showing comb-cracks on scales large and small. (a) Satellite image of the sea ice cover off the coast of Greenland, 25 April 1976 (from Erlingsson 1988). (b) Suggested shear loading of the macrocolumns created by the secondary cracks. (c) Cracks in fresh-water granular ice of ~10 mm grain size loaded triaxially acrosscolumn (R ~ 0.05) at -40°C at 10-3 s-1.

and compressive shear faults has already been noticed. Figure 27 shows another example; namely, comb-like cracks in the Greenland Sea ice field and in a small specimen of granular ice. These “look-alike” features heighten the possibility that the failure mechanisms may be scale-independent. Indeed, recent fractal analyses of crack patterns accompanying the fracture and fragmentation of ice (Weiss 2001; Weiss and Gay 1988) revealed no characteristic scale in over nine orders of magnitude of length (10-4 m to

 

Brittle Failure of Ice 

245 

105 m), except one corresponding to the mean grain size during the early stage of deformation. This result precludes the existence of an elementary volume within which damage is representative of the ice as a whole (Weiss 2001), and suggests universal scaling (Weiss and Gay 1998) of the fracture processes. Thus, if the wing-crack mechanism, which is scale-independent, accounts for the in-field wing-like leads of Figure 23 ( l a ≈ 5; a ≈ 40 km), and if there are no major effects of scale on the values of the physical parameters (KIc = 0.25 Mpa√m (Dempsey 1996); μ = 0.5), then from the non-dimensional quantification of that mechanism (Fig. 13b) those leads could have formed under compressive stresses of ~10 kPa to 30 kPa. Similarly, if the comb-crack mechanism, also scale-independent, accounts for the large-scale features of Figure 27 [h/w ≈ 3; Kc = 0.25 Mpa√m; µ = 0.5], then from Equation (17) those leads could have formed under in-field compressive stresses of ~3 kPa. Interestingly, these estimates are of the same order of magnitude as peak compressive stresses measured in pack ice, which typically range from 10 kPa to 100 kPa (Coon et al. 1989; Tucker and Perovich 1992; Richter-Menge and Elder 1998). Unfortunately, unlike the laboratory, the field has not yet offered a direct comparison between the peak stresses and the underlying mechanical events. Implicit in the discussion so far is the idea that the arctic sea ice cover behaves in a macroscopically brittle manner. Given that covers typically deform at low strain rates, ~1%/day (10-7 s-1), brittle behavior might appear to contradict the analysis presented above (in the section Ductile-to-Brittle Transition Under Compression), where the ductile-to-brittle transition strain rate for salt-water ice loaded across the columns was noted to be ~10-3 s-1. However, the earlier discussion focused on the behavior of small pieces of material. The ice cover on the Arctic ocean is a very large body, thousands of kilometers wide and up to ~3 m thick: it contains stress concentrators of length λ— thermal cracks, for instance—that can be several orders of magnitude larger than cracked grain boundaries. In terms of the above analysis, in which the transition strain rate scales as λ-1.5, one need only imagine in-field concentrators longer than a few meters to account for brittle behavior at strain rates ≤10-7 s-1. Another point (in relation to the ductile-brittle transition) is worth noting. In modeling ice sheet deformation, it is common practice to view the cover as a continuum, albeit an anisotropically damaged one. Hibler (2001) develops this theme. Invoked there is the notion of viscous-plastic deformation. “Viscous behavior” within that context has the same meaning as creep-based, ductile behavior within the present context; and “plastic behavior” has the same meaning as brittle failure used here. In other words, within the context of ice sheet modeling, the transition from viscous to plastic deformation is essentially the ductile-to-brittle transition. To summarize, the evidence reveals a striking similarity in the pattern of cracking on widely different spatial scales. This suggests that the failure mechanisms are scaleindependent. Owing to the concentration of stress by internal flaws of lengths, λ, the applied stress to operate the mechanisms appears to scale roughly as λ-0.5. However, before firm conclusions can be drawn about scale-independence, more work is needed to relate in-field stress amplitudes to specific physical events. ICE AND ROCK Before closing, it is worth noting again that many of the phenomena exhibited by ice—e.g., pressure-sensitive deformation, ductile-to-brittle transition, and compressive shear faulting on scales small and large—are displayed by rock (see e.g., Duba et al. 1990, and references therein). Also, the fracture features that accompany compressive failure—distributed cracks, wing cracks and comb cracks—are also seen in geomaterials.

246 

Schulson 

And the evolution of a shear fault in granite (Lockner et al. 1991), for instance, is remarkably similar to its evolution in ice, even to the point of initiating at a free surface and then propagating along a narrow band of damage inclined by about 30° to the direction of the most compressive principal stress. These similarities, as already noted, strongly imply that failure mechanisms are common to many brittle or semi-brittle materials (Renshaw and Schulson 2001). Also common may be the criterion for the ductile-to-brittle transition, described by Equation (21). In other words, lessons learned from a study of ice may help to elucidate the failure of rock, and vice versa. ACKNOWLEDGMENTS

The author acknowledges Profs. Ali Argon, Ian Baker, Bill Hibler, Harold J. Frost, Mark Kachanov, Carl Renshaw and Kaj Riska and Drs. Jerome Weiss, Paul Duval, Peter Sammonds, Devinder Sodhi, Ron Kwok, Don Nevel and Chris Heuer for helpful comments. He also recognizes again his former students whose published work has greatly contributed to this story: Rachel Batto, Stephanie Buck, Neil Cannon, Michele Couture, John Currier, Doug Fifolt, Mark Gies, Eric Gratz, Rebecca Haerle, Steven Hoxie, Daniel Iliescu, Doug Jones, Gary Kuehn, Greg Lasonde, Russell Lee, Pooi Lim, Jeff Melton, Oleg Nickolayev, Suogen Qi and Timothy Smith. Also, he acknowledges Drs. Ron Kwok and Harry Sterns for copies of Figure 25, and Dr. James Wilson for pointing out these images in the first place, following a presentation by the author. Finally, he gratefully acknowledges the J. William Fulbright Foreign Scholarship Board for a research award. This work was supported by the Army Research Office, the Office of Naval Research, National Oceanographic and Atmospheric Administration, and Exxon and Mobil. REFERENCES Ahmad S, Whitworth RW (1988) Dislocation-motion in ice–a study by synchrotron X-ray topography. Phil Mag A 57:749-766 Anderson DL, Weeks WF (1958) A theoretical analysis of sea ice strength. Trans Am Geophys Union 39:632-640 API (1993) Planning, designing, constructing structures, pipelines for arctic conditions. American Petroleum Institute, Dallas, Texas Ashby MF, Hallam SD (1986) The failure of brittle solids containing small cracks under compressive stress states. Acta Metall 34:497-510 Assur A (1958) Composition of sea ice, its tensile strength. In Arctic Sea Ice. U S Natl Acad Sci, Natl Res Council, Washington, DC, p 106-138 Baker I (1997) Observation of dislocations in ice. J Phys Chem B 101:6158-6162 Batto RA, Schulson EM (1992) A preliminary investigation of the ductile-brittle transition in columnar S2 ice under compression. IAHR Ice Symp 1021-1034 Batto RA, Schulson EM (1993) On the ductile-to-brittle transition in ice under compression. Acta metall mater 41:2219-2225 Bazant ZP, Li Y-N, Jirásek M, Li Z, Kim J-J (1995) Effect of size on distributed damage, fracture of sea ice. Northwest Research Associates, Inc., Bellevue, Washington: Sea Ice Mechanics, Arctic Modeling Workshop 1:73-83 Brace WF, Bombolakis EG (1963) A note of brittle crack growth in compression. J Geophys Res 68:3709 Buck SE, Nickolayev OY, Schulson EM (1994) Failure envelopes, the ductile to brittle transition in S2 columnar freshwater, saline ice under biaxial compression. IAHR 94, 12th Intl Symp on Ice 454-463 Butkovich TR (1954) Ultimate strength of ice. U S Snow, Ice, Permafrost Research Establishment, Res Paper 15 Campbell J, Fahmy Y, Conrad H (1999) Plastic deformation kinetics of fine-grained alumina. Metall Mater Trans A 30A:2809-2816 Cannon NP, Schulson EM, Smith TR, Frost HJ (1990) Wing cracks, brittle compressive fracture. Acta metall mater 38:1955-1962

 

Brittle Failure of Ice 

247 

Carter D (1971) Lois et mechanismes de l'apparente fracture fragile de la glace de riviere et de lac. University of Laval, France Castelnau O, Canova GR, Lebensohn RA, Duval P (1997) Modeling viscoplastic behavior of anisotropic polycrystalline ice with a self-consistent approach. Acta Mater 45:4823-4834 Chantikul P, Bennison SJ, Lawn BR (1990) Role of grain size in the strength, r-curve properties of alumina. J Am Ceram Soc 73:2419-2427 Cole DM (1985) Grain size, the compressive strength of ice. Am Soc Mech Engin, 4th Intl Symp Offshore Mechanics, Arctic Engin, p 220 Cole DM (1987) Strain rate, grain size effects in ice. J Glaciol 33:274-280 Cole DM (1990) Reversed direct-stress testing of ice: Initial experimental results, analysis. Cold Regions Sci Techn 18:303-321 Cole DM (2001) The microstructure of ice, its influence on mechanical properties. Engin Fracture Mech 68:1797-1822 Conrad RE, II, Friedman M (1976) Microscopic feather fractures in the faulting process. Tectonophysics 33:187-198 Cooke ML (1997) Fracture localization along faults with spatially varying friction. J Geophys Res 102:24425-24434 Coon MD, Knoke GS, Echert DC, Pritchard RS (1989) The architecture of an anisotropic elastic-plastic sea ice mechanics constitutive law. J Geophys Res 103:21915-21925 Costin LS (1983) A microcrack model for the deformation, failure of brittle rock. J Geophys Res 88: 9485-9492 Cottrell AH (1958) Theory of brittle fracture in steel, similar metals. Trans Metall Soc AIME 212:192-203 Couture ML, Schulson EM (1994) The cracking of ice under rapid unloading. Phil Mag Lett 69:9-14 Cruikshank KM, Zhao G (1991) Analysis of minor fractures associated with joints, faulted joints. J Struct Geol 13:865-886 Currier JH, Schulson EM (1982) The tensile strength of ice as a function of grain size. Acta Metall 30:1511-1514 Davies RK, Pollard DD (1986) Relations between left-lateral strike-slip faults, right-lateral kink bands in granodiorite, Mt. Abbot Quadrangle, Sierra Nevada, California. Pure Appl Geophys 124:177-201 Dempsey JP (1991) Fracture toughness of ice. In Ice Structure Interactions. SJ Jones et al. (eds) SpringerVerlag, Berlin, p 109-125 Dempsey JP (1996) Scale effects on the fracture of ice. The Minerals, Metals, Materials Society, The Johannes Weertman Symposium 19:351-361 Dempsey JP, Adamson RM, Mulmule SV (1999) Scale effects on the in situ tensile strength, fracture of ice. Part ii: First-year sea ice at resolute, N.W.T. Intl J Fracture 95:347-366 Duba AG, Durham WB, Handin JW, Wang HF (eds) (1990) The brittle-ductile transition in rocks: The Heard Volume. Geophys Monogr 56, Am Geophys Union, Washington, DC Durham WB, L.A.Stern (2001) Rheological properties of water ice, Annual Rev. Earth Planet. Science 29:295-330 Duval P, Ashby MF, Anderman I (1983) Rate-controlling processes in the creep of polycrystalline ice. J Phys Chem 87:4066-4074 Duval P, Arnaud L, Brissaud O, Montagnat M, de la Chapelle S (2000) Deformation, recrystallization processes of ice from polar ice sheets. Ann Glaciol 30:83-87 Dykins JE (1970) Ice engineering: Tensile properties of sea ice grown in a confined system. Naval Civil Engineering Laboratory, Port Hueneme, California Erlingsson B (1988) Two-dimensional deformation patterns in sea ice. J Glaciol 34:301-308 Evans AG, Rice JR, Hirth JP (1980) Suppression of cavity formation in ceramics: Prospects for superplasticity. J Am Ceram Soc 63:368-375 Flemings MC (eds) (1974) Solidification Processing. McGraw-Hill, New York Fletcher NH (1973) The surface of ice. In Physics, Chemistry of Ice. Whalley E, Jones SJ, Gold LW (ed) Royal Soc of Canada, Ottawa, p 132-136 Frederking R (1977) Plane-strain compressive strength of columnar-grained, granular-snow ice. J Glaciol 18:505-516 Frost HJ (2001) Crack nucleation in ice. Engin Fracture Mechanics 68:1823-1837 Gagnon RE, Gammon PH (1995) Triaxial experiments on iceberg, glacier ice. J Glaciol 41:528-540 Gammon PH, Kiefte H, Clouter MJ, Denner WW (1983) Elastic constants of artificial ice, natural ice samples by brillouin spectroscopy. J Glaciol 29:433-460 Germanovich LN, Salganik RL, Dyskin AV, Lee KK (1994) Mechanisms of brittle failure of rock with preexisting cracks in compression. Pure Appl Geophys 19:197-212 Gies MC (1988) The development of brush-type platens for loading ice specimens in compression. Master of Engineering thesis, Dartmouth College, Hanover, New Hampshire

248 

Schulson 

Glen JW (1968) The effect of hydrogen disorder on dislocation movement, plastic deformation in ice. Phys kondens Mater 7:43-51 Gold LW (1997) Statistical characteristics for the type, length of deformation-induced cracks in columnargrain ice. J Glaciol 43:311-320 Gottschalk RR, Kronenberg AK, Russel JE, Handin J (1990) Mechanical anisotropy of gniess: Failure criterion, textural sources of directional behavior. J Geophys Res 95:613-621 Granier T (1985) Origin, damping, pattern development of faults in granite. Tectonics 4:721-737 Grape JA, Schulson EM (1992) Effect of confining stress on brittle indentation failure of columnar ice. J Offshore, Polar Engineering 2:212-221 Gratz ET, Schulson EM (1997) Brittle failure of columnar saline ice under triaxial compression. J Geophys Res 102:5091-5107 Greeley R, Figueredo PH, Williams DA et al. (2000) Geologic mapping of Europa. J Geophys Res 105:22559-22578 Greenberg R, Geissler P, Tufts BR, Hoppa GV (2000) Habitability of Europa's crust: the role of tidaltectonic processes. J Geophys Res 105:17551-17562 Greenberg RP, Weidenschilling SJ (1984) How fast do Galilean satellites spin? Icarus 58:186-196 Gupta V, Bergström JS (1998) Compressive failure of rocks by shear faulting. J Geophys Res 103: 23875-23895 Hallam SD, Duval P, Ashby MF (1987) A study of cracks in polycrystalline ice under uniaxial compression. J Physique 48:303-311 Hausler FU (1981) Multiaxial compressive strength tests on saline ice using brush-type loading platens. IAHR Ice Symp, p 389-398 Hawkes I, Mellor M (1972) Deformation, fracture of ice under uniaxial stress. J Glaciol 11:103-131 Haynes FD, Mellor M (1977) Measuring the uniaxial compressive strength of ice. J Glaciol 19:213-223 Hibler WD, Schulson EM (2000) On modeling the anisotropic failure, flow of flawed sea ice. J Geophys Res 105:17105-17120 Hibler WDI (2001) Sea ice fracturing on the large scale. Engineering Fracture Mechanics 68:2013-2044 Hill R (1950) The Mathematical Theory of Plasticity. Oxford University Press, New York Hobbs DW, Pomeroy CD, Newman JB (1977) Design stresses for concrete structures subject to multi-axial stresses. Struct Engin 55:151-164 Hopkins MA (1998) Four stages of pressure ridging. J Geophys Res 103:21883-21891 Hoppa GV, Tufts BR, Greenberg R, Geissler PE (1999) Formation of cycloidal features on Europa. Science 285:1899-1902 Horii H, Nemat-Nasser S (1985) Compression-induced microcrack growth in brittle solids: Axial splitting, shear failure. J Geophys Res 90:3105-3125 Horii H, Nemat-Nasser S (1986) Brittle failure in compression: Splitting, faulting, brittle-ductile transition. Phil Trans R Soc A319:337-374 Hu X, Jia K, Baker I, Black D (1995) Dislocation mobility in HCl-doped ice. Mater Res Soc 375:287-292 Hughes T (1989) Calving ice walls. Ann Glaciol 12:74-80 Iliescu D, Schulson EM (1998) Macroscopic compressive shear faults in S2 columnar ice. A.A. Balkema Publishers, 14th Intl Symp on Ice, p 553-558 Ingraffea AR (1981) Mixed-mode fracture initiation in Indiana limestone, Westerly granite. Proc U S Symp Rock Mechanics 22:196-191 Jaeger JC, Cook NGW (1979) Fundamentals of Rock Mechanics. Chapman-Hall, London Jellinek HHG (1957) Tensile strength properties of ice adhering to stainless steel. U S Snow, Ice, Permafrost Research Establishment Research Report 23 Johanessen OM, Miles MW, Bjorgo E (1995) The arctic shrinking sea-ice. Nature 376:126-127 Johnson KL (1987) Contact mechanics. Press Syndicate of the University of Cambridge, Cambridge, UK Jones SJ (1982) The confined compressive strength of polycrystalline ice. J Glaciol 28:171-177 Jordaan IJ (2001) Mechanics of ice-structure interaction. Engineering Fracture Mechanics 68:1923-1960 Kachanov ML (1982) A microcrack model of rock inelasticity, part II: Propagation of microcracks. Mech Mater 1:29-41 Kalifa P, Duval P, Ricard M (1989) Crack nucleation in polycrystalline ice under compressive stress states. American Society of Mechanical Engineers, New York, Eighth Intl Conference on Offshore Mechanics: Arctic Engineering 4:13-20 Karge JS, Kaye JZ, Head (III) JW, Marion GM, Sassen R, Crowley JK, Ballseteros OP, Grant SA, Hogenboom DL (2000) Europa's crust and ocean: origin, composition, and the prospects for life. Icarus 148:226-265 Kemeny JM, Cook NGW (1987) Crack models for the failure of rock under compression. 2nd Intl Conf Constitutive Laws for Engineering Materials 2:879-889

 

Brittle Failure of Ice 

249 

Kennedy FE, Schulson EM, Jones D (2000) Friction of ice on ice at low sliding velocities. Phil Mag A 80:1693-1110 Ketcham WM, Hobbs PV (1969) An experimental determination of the surface energies of ice. Phil Mag 19:1161-1173 Khomichevskaya IS (1940) O vremennom soprotivlenii szhatiyu vechnomerzlykh gruntov i l'da yestestvennoy struktury. (The ultimate compressive strength of permafrost, ice in their natural states). Trudy Komiteta po Vechnoy Merzlote 10:37-83 Knott JF (1973) Fundamentals of Fracture Mechanics. John Wiley & Sons, New York Korzhavin KN (1955) Vliyaniye skorosti deformirovaniya na velichinu predela prochnosti rechnogo l'da priodnoosnom szhatii [the effect of the speed of deformation on the ultimate strength of river ice subject to uniaxial compression]. Trudy Novosibirskogo Instituta Inzhenerov Zheleznodorozhnogo Transporta 11:205-216 Kuehn G, Schulson E, Jones D, Zhang J (1993a) The compressive strength of ice cubes of different sizes. Transactions of the ASME, 11th Intl Symp, Exhibit on Offshore Mechanics, Arctic Engineering, p 142-148 Kuehn GA, Schulson EM (1994a) The mechanical properties of saline ice under uniaxial compression. Ann Glaciol 19:39-48 Kuehn GA, Schulson EM (1994b) Ductile saline ice. J Glaciol 40:566-568 Kuehn GA, Schulson EM, Nixon WA (1988a) The effect of pre-strain on the compressive ductile-to-brittle transition in ice. IAHR Symp 1988 Sapporo, p 109-117 Kuehn GA, Schulson EM, Nixon WA (1988b) The effects of pre-strain on the compressive ductile-tobrittle transition in ice. IAHR Ice Symp, p 109-117 Kuehn GA, Schulson EM, Jones DE, Zhang J (1993b) The compressive strength of ice cubes of different sizes. Transactions of the ASME, 11th Intl Symp, Exhibit on Offshore Mechanics, Arctic Engineering 115:142-148 Kwok R, Rignot E, Holt B, Onstott R (1992) Identification of sea ice types in space-borne synthetic aperture radar data. J Geophys Res-Oceans 97:2391-2402 Kwok RE, Cunningham GF, LaBelle-Hamer N, Holt B, Rothrock D (1999). Ice thickness derived from high-resolution radar imagery. EOS Trans Am Geophys Union, October 19, 1999, p 494-497 Lee RW, Schulson EM (1988) The strength, ductility of ice under tension. J Offshore Mechanics, Arctic Engin 110:187-191 Lehner F, Kachanov M (1996) On modelling of "winged" cracks forming under compression. Intl J Fracture 77:R69-75 Lewis JK (1995) A conceptual model of the impact of flows on the stress state of sea ice. J Geophys Res 100:8819-8825 Lindsay RW, Rothrock DA (1995) Arctic sea-ice leads from advanced very high-resolution radiometer images. J Geophys Res–Oceans 10:4533-4544 Liu F, Baker I (1995) Dislocation-grain boundary interactions in ice crystals. Phil Mag A 71:15-42 Lockner DA, Byerlee JD, Kuksenko V, Pnomarev A, Sidorin A (1991) Quasi-statis fault growth, shear fracture energy in granite. Nature 350:39-42 Manley ME, Schulson EM (1997) On the strain-rate sensitivity of columnar ice. J Glaciol 43:408-410 Marko JR, Thomson RE (1977) Rectilinear leads, internal motions in the ice pack of the western arctic ocean. J Geophys Res 82:979-987 Martel SJ, Pollard DD, Segall P (1988) Development of simple strike-slip fault zones in granitic rock, Mount Abbott Quadrangle, Sierra Nevada, California. Geol Soc Am Bull 100:1451-1465 Martel SM, Pollard DD (1989) Mechanics of slip, fracture along small faults, simple strike-slip fault zones in granitic rock. J Geophys Res–Solid Earth, Planets 94:9417-9428 Maykut GA (1982) Large-scale heat exchange, ice production in the central arctic. J Geophys Res 87: 7971-7984 McClintock FA, Walsh JB (1962) Frictions in Griffith’s cracks in rock under pressure. Proc 4th U S Natl Congress Appl Mech, p 1015 Melton JS, Schulson EM (1998) Ductile compressive failure of columnar saline ice under triaxial loading. J Geophys Res 103:21759-21766 Michel B (1978) Ice Mechanics. Laval University Press, Quebec Moore DE, Lockner DA (1995) The role of microcracking in shear-fracture propagation in granite. J Struct Geol 17:95-114 Nickolayev OY, Schulson EM (1993) Failure envelopes, the ductile to brittle transition in columnar saline ice under biaxial compression. 1st Joint Mechanics Meeting of ASME, ASCE, SES Meeting '93 163:61-69 Nickolayev OY, Schulson EM (1995) Grain-boundary sliding, across-column cracking in columnar ice. Phil Mag Letters 72:93-97

250 

Schulson 

Nixon WA, Schulson EM (1987) A micromechanical view of the fracture toughness of ice. J Physique 48:313-319 Nixon WA, Schulson EM (1988) Fracture toughness of ice over a range of grain sizes. J Offshore Mech Arctic Eng 110:192-196 Overland JE, McNutt SL, Salo S, Groves J, Li SS (1998) Arctic sea ice as a granular plastic. J Geophys Res–Oceans 103:21845-21867 Pappalardo, RT, Belton MJS, Breneman HH, et al. (1999) Does Europa have a subsurface ocean? Evaluation of the geological evidence. J Geophys Res 104:24015-24055 Paterson MS (1978) Experimental Rock Deformation—The Brittle Field. Springer-Verlag, New York Patterson WSB (1994) The physics of glaciers. Oxford University Press, Oxford, UK Petit J-P, Barquins M (1988) Can natural faults propagate under mode ii conditions? Tectonics 7: 1243-1256 Peyton HR (1966) Sea Ice Strength. University of Alaska, Geophysical Institute, Fairbanks Picu RC, Gupta V (1995a) Observations of crack nucleation in columnar ice due to grain boundary sliding. Acta metall mater 43:3791-3797 Picu RC, Gupta V (1995b) Crack nucleation in columnar ice due to elastic anistropy, grain boundary sliding. Acta metall mater 43:3783-3789 Pradeau SI, Schulson EM (1998) Failure of columnar saline ice under biaxial compression at -2.5°C (unpublished). Thayer School of Engineering, Dartmouth College, Hanover, New Hampshire Qi S, Schulson EM (1998) The effect of temperature on the ductile-to-brittle transition in columnar ice. A.A. Balkema Publishers, 14th Intl Symp on Ice, p 521-527 Reches Z, Lockner DA (1994) Nucleation, growth of faults in brittle rocks. J Geophys Res 99: 18159-18173 Renshaw CE, Schulson EM (1998) Non-linear rate dependent deformation under compression due to state variable friction. Geophys Res Letters 25:2205-2208 Renshaw CE, E.M.Schulson (2001) Universal behavior in compressive failure of brittle materials. Nature 412: 897-900 Reynolds RT, McKay CP, Casting JF (1987) Europa, tidally heated oceans, and habitable zones around giant planets. Adv Space Res 7:5125-5132 Richter-Menge J (1991) Confined compressive strength of horizontal first-year sea ice samples. J Offshore Mech Arctic Engin 113:344-351 Richter-Menge JA, Jones KF (1993) The tensile strength of first-year sea ice. J Glaciol 39:609-618 Richter-Menge JA, Elder BC (1998) Characteristics of pack ice stress in the Alaska’s Beaufort Sea. J Geophys Res–Oceans 103:21817-21829 Riedel H, Rice JR (1980) Tensile cracks in creeping solids. ASTM-STP 7700:112-130 Riska K, Tuhkuri J (1995) Application of ice cover mechanics in design, operations of marine structures. Northwest Research Associates, Inc., Bellevue, Washington: Sea Ice Mechanics, Arctic Modeling Workshop 2:123-134 Rispoli R (1981) Stress fields about strike-slip faults from stylolites, tension gashes. Tectonophysics 75:T29-T36 Rist MA (1997) High stress ice fracture, friction. J Phys Chem B 101:6263-6266 Rist MA, Murrell SAF (1994) Ice triaxial deformation, fracture. J Glaciol 40:305-318 Rist MA, Jones SJ, Slade TD (1994) Microcracking, shear fracture in ice. Ann Glaciol 19:131-137 Rist MA, Sammonds PR, Murrell SAF, Meredith PG, Doake CSM, Oerter H, Matsuki K (1999) Experimental, theoretical fracture mechanics applied to antarctic ice fracture, surface crevassing. J Geophys Res 104:2973-2987 Rudnicki JW, Rice JR (1975) Conditions for the localization of deformation in pressure-sensitive dilatant materials. J Mech Phys Solids 23:371-394 Sammonds PR, Murrell SAF, Rist MA (1998) Fracture of multi-year sea ice. J Geophys Res 100: 21795-21815 Sanderson TJO (1988) Ice mechanics: Risks to offshore structures. Graham & Trotman, London Schulson EM (1979) An analysis of the brittle to ductile transition in polycrystalline ice under tension. Cold Regions Sci Techn 1:87-91 Schulson EM (1987) The fracture of ice 1h. J Physique C1-207–C-220 Schulson EM (1990) The brittle compressive fracture of ice. Acta metall mater 38:1963-1976 Schulson EM (1996) The failure of ice under compression. TMS Annual Meeting, Weertman Symposium, p 363-374 Schulson E.M (2001) Fracture of ice. Engg Fracture Mech 68:1793-1795 Schulson E.M (2002) Compressive shear faults within the arctic sea ice cover. J Geophys Res (submitted) Schulson EM (2002 On compressive shear faulting in ice: Coulombic vs. plastic faults. Acta Mater 50:3415-3424

 

Brittle Failure of Ice 

251 

Schulson EM, Cannon NP (1984) The effect of grain size on the compressive strength of ice. IAHR Ice Symp Schulson EM, Hibler WD (1991) The fracture of ice on scales large, small: Arctic leads, wing cracks. J Glaciol 37(127):319-323 Schulson EM, Kuehn GA (1993) Ductile ice. Phil Mag Lett 67:151-157 Schulson EM, Buck SE (1995) The ductile-to-brittle transition, ductile failure envelopes of orthotropic ice under biaxial compression. Acta metall mater 43:3661-3668 Schulson EM, Nickolayev OY (1995) Failure of columnar saline ice under biaxial compression: Failure envelopes, the brittle-to-ductile transition. J Geophys Res 100:22383-22400 Schulson EM, Gratz ET (1999) The brittle compressive failure of orthotropic ice under triaxial loading. Acta Mater 47:745-755 Schulson EM, Lim PN, Lee RW (1984) A brittle to ductile transition in ice under tension. Phil Mag A 49:353-363 Schulson EM, Hoxie SG, Nixon WA (1989a) The tensile strength of cracked ice. Phil Mag A 59: 303-311 Schulson EM, Jones DE, Kuehn GA (1991a) The effect of confinement on the brittle compressive fracture of ice. Ann Glaciol 15:216-221 Schulson EM, Iliescu D, Renshaw CE (1999) On the initiation of shear faults during brittle compressive failure: A new mechanism. J Geophys Res 104:695-705 Schulson EM, Geis MC, Lasonde GJ, Nixon WA (1989b) The effect of the specimen-platen interface on the internal cracking, brittle fracture of ice under compression: High-speed photography. J Glaciol 35:378-382 Schulson EM, Kuehn GA, Jones DE, Fifolt DA (1991b) The growth of wing cracks, the brittle compressive failure of ice. Acta metall mater 39:2651-2655 Schulson EM, Gratz ET, Johnson RC, Spring W (unpublished research) Schulson EM, Baker I, Robertson CD, Bolon RB, Harnimon RJ (1989c) Fractography of ice. J Mater Sci Lett 8:1193-1194 Schwarz J, Weeks WF (1977) Engineering properties of sea ice. J Glaciol 19:499-531 Segall P, Pollard DD (1983) Nucleation, growth of strike-slip faults in granite. J Geophys Res 88:555-568 Shapiro LH, Weeks WF (1993) The influence of crystallographic, structural properties on the flexural strength of small sea ice beam. Am Soc Mech Eng, Ice Mechanics–1993; 1993 Joint ASME Applied Mechanics, Materials Summer Mtg, Appl Mech Div 163:18159-18173 Shearwood C, Whitworth RW (1991) The velocity of dislocations in ice. Phil Mag A 64:289-302 Shen W, Lin SZ (1986) Fracture toughness of bohai bay sea ice. 5th Intl Offshore Mech, Arctic Engin Symp IV:354-357 Sinha NK (1982) Constant strain-, stress-rate compressive strength of columnar-grained ice. J Mater Sci 17:785-802 Smith E, Barnby JT (1967) Crack nucleation in crystalline solids. Metall Sci J 1:56-64 Smith TR, Schulson EM (1993) The brittle compressive failure of fresh-water columnar ice under biaxial loading. Acta metall mater 41:153-163 Smith TR, Schulson EM (1994) Brittle compressive failure of salt-water columnar ice under biaxial loading. J Glaciol 40:265-276 Sodhi DS (2001) Crushing failure during ice-structure interaction. Eng Fracture Mechanics 68:1889-1921 Thouless MD, Evans AG, Ashby MF, Hutchinson JW (1987) The edge cracking, spalling of brittle plates. Acta metall 35:1333-1341 Timco GW, Frederking RMW (1983) Confined compressive strength of sea ice. 7th Intl Conf, Port and Ocean Engineering Under Arctic Conditions (POAC) I:243-253 Trickett YL, Baker I, Pradhan PMS (2000) The orientation dependence of the strength of ice single crystals. J Glaciol 46:41-44 Tucker WB, Perovich DK (1992) Stress measurements in drifting pack ice. Cold Regions Sci Techn 20: 119-139 Voytkovskiy KF (1960) Mekhanicheskiye svaystva l'da [mechanical properties of ice]. Moscow, Izdatel'stvo Akademii Nauk SSSR, [U S Dept of Commerce, Office of Technical Services: transl] Weeks WF (1962) Tensile strength of nacl ice. J Glaciol 4:25-52 Weeks WF, Ackley SF (1982) The growth, structure, properties of sea ice. U S Army Cold Regions Research, Engineering Laboratory, November 1982 Weertman J (1983) Creep deformation of ice. Ann Rev Earth Planet Sci 11:215-240 Weiss J (1999) The ductile behaviour of damaged ice under compression. Port and Ocean Engineering Under Arctic Conditions (POAC), p 10 Weiss J (2001) Fracture, fragmentation of ice: A fractral analysis of scale invariance. Engin Fracture Mech (in press)

252 

Schulson 

Weiss J, Schulson EM (1995) The failure of fresh-water granular ice under multi-axial compressive loading. Acta metall mater 43:2303-2315 Weiss J, Gay M (1998) Fracturing of ice under compression creep as revealed by a multifractal analysis. J Geophys Res 103:24005-24016 Weiss J, Schulson EM (2000) Grain boundary sliding, crack nucleation in ice. Phil Mag 80:279-300 Weiss J, Schulson EM, Frost HJ (1996) The nucleation of microcracks in ice cubes compressed equally on all boundaries. Phil Mag A 73:1385-1400 Wong T-F (1982) Micromechanics of faulting in Westerly granite. Intl J Rock Mech Mining Sci 19:49-64 Zimmermann A, Rodel J (1998) Generalized Orowan-Petch plot for brittle fracture. J Am Ceram Soc 81:2527-2532

9

Seismic Wave Attenuation: Energy Dissipation in Viscoelastic Crystalline Solids Reid F. Cooper Department of Materials Science and Engineering University of Wisconsin–Madison Madison, Wisconsin 53706 INTRODUCTION

Seismic imaging as a tool to understand the structure and, perhaps, the dynamics of the planet at depth involves the spatially resolved, combined study of wave velocities and velocity dispersion, wave birefringence and wave attenuation (mechanical absorption). When integrated with insights from petrology, plus-or-minus input from magnetotellurics, specific hypotheses concerning structure at depth can be formulated (cf. Karato 1993; Karato and Karki 2001). The recent advances in tomography have allowed significant improvements in spatial resolution which also have allowed ever more exacting hypotheses of structure to be articulated. Nevertheless, as is nicely illustrated by the relatively recent seismic analyses of an accreting plate margin (the East Pacific Rise —e.g., Toomey et al. 1998; Webb and Forsyth 1998), the structures inferred at depth can vary significantly: clearly, the interpretation of seismic data is limited specifically by a lack of understanding of the physical processes by which low-frequency wave absorption occurs, particularly in the cases (a) where melt is present and/or (b) where the material is being actively plastically deformed. This chapter emphasizes the mineral physics/materials science of mechanical absorption, specifically in dense materials at elevated temperature; the interest, then, is in employing the physical study of mechanical absorption to understand natural phenomena such as those mentioned above. Beyond questions of the basic mineral physics of absorption, the guiding interest the community of scholars pursuing the ideas is in isolating actual absorption mechanisms operative in the geological setting thereby allowing for a greater discrimination in the interpretation of seismic data. At the outset, I note that this contribution does not seek to be a comprehensive review of the ideas and work involved in the physics of mechanical absorption. There are marvelous reviews in the literature that are both authoritative and current; I refer to these frequently in contemplating my own experimental work and so am pleased to point them out here. In the geophysics literature, Karato and Spetzler (1990) have extensively reviewed the variety of lattice-defect-based mechanical loss mechanisms described theoretically in the physics and materials science communities, with an eye towards their application to minerals, while Jackson (1993, 2000) reviewed experimental approaches geared to study these effects in mineral systems. In the materials science literature, the text by Nowick and Berry (1972) remains the standard. In the mechanics literature, the books by Findley et al. (1976; corrected and republished 1989) and, recently, by Lakes (1999) accurately tie down the phenomenology of attenuation. Additionally, in that all of the above authors point out the importance of grain boundaries and solid-state phase boundaries in effecting attenuation (at least in part), Kê’s (1999) recent review of a halfcentury of his own work in the area is significant. LINEAR VISCOELASTICITY: A CHEMICAL KINETICS PERSPECTIVE The mechanical properties of solids are distinctly time dependent. When a deviatoric-stress thermodynamic potential is applied to a solid system initially at 1529-6466/00/0051-0009$05.00

254

Cooper

mechanical and chemical equilibrium, that system responds—specifically in the form of strain—in a manner that minimizes the new potential: LeChatlier’s principle holds in mechanical properties as well as in thermochemistry. (From this point forward, I will refer to the deviatoric stress potential as the “stress potential” or simply the “stress;” “mechanical potential” is used in most thermodynamics treatises to describe pressure, which, of course, must incorporate the hydrostatic component of the applied tensorial stress. Further, hydrostatic pressure can effect mechanical absorption, too: the meltmigration absorption response of partial melts being one example.) The responses to the stress potential combine elastic and inelastic phenomena; the resultant behavior overall is labeled “viscoelasticity.” Elastic response, i.e., strain in the form of the stretching and bending of atomic bonds, is considered instantaneous or time-independent (to first order, i.e., ignoring inertial effects). In contrast, inelastic responses are distinctly time dependent: in fully crystalline solids, they involve the interaction of the stress potential with point (vacancies, interstitials, substitutionals, as well as jogs and kinks on dislocations), line (dislocations) and planar (twin boundaries; grain and phase boundaries) lattice defects. Time dependence is a manifestation that the strain so produced involves the motion of these lattice defects: the motion involves the breaking and reforming of atomic bonds, the rate of which is dictated by Boltzmann statistics (and, thus, the rate depends exponentially on temperature). One can further divide inelastic responses, identifying those that produce time-dependent but fully recoverable strain (“anelastic”) and those that produce time-dependent, permanent (irrecoverable, “plastic”) strain. In that viscoelasticity combines elastic and inelastic responses, viscoelasticity is time dependent. From a thermodynamics perspective, one understands elasticity as a reversible process where strain energy (i.e., the energy manifest as the distortion of bonds) is stored in the material and is recoverable without loss, while inelastic response is irreversible: strain energy is dissipated within the material, that is, it is converted irreversibly to heat. (Of course, if fracture is initiated, strain-energy dissipation sees the formation of new surfaces as well as the conversion to heat. This chapter, though, will not consider fracture behavior.) Kinetics enters the discussion—and dominates much of the consideration of mechanical response—when (i) temperature is sufficiently high and/or (ii) the application of a deviatoric stress sufficiently slow and/or (iii) the deviatoric stress persists for a sufficiently long time such that either (a) the inelastic strain response(s) occurs at a rate(s) competitive to the elastic response or (b) the accumulated strain produced by the inelastic responses outstrips that from the elastic response. The elastic and many inelastic responses of materials act in “parallel” kinetically, meaning the responses to stress are independent physically and, as such, the strain produced by each response is additive. Inelastic responses, however, often involve a number of “serial” kinetic steps, which are physical processes that must act in sequence in order to produce strain. Some series-process phenomena are inherent in the process physics (e.g., solution-precipitation—or pressure-solution—creep involves the serial physical steps of (i) atomic dissolution into the liquid phase, (ii) atomic/ionic transport through the liquid phase and (iii) precipitation from the liquid phase; dislocation creep involves the serial process of climb (or cross-slip) and glide of the lattice dislocations), while others are based on geometrical constraints (e.g., diffusional creep acts in series with the processes of grain boundary sliding). The distinction between parallel and serial kinetic processes in mechanical response is important: for a given set of thermodynamic potentials—pressure, temperature, various chemical potentials (e.g., water fugacity, oxygen fugacity, oxide chemical potentials, etc.), and deviatoric stress—and microstructure parameters (e.g., grain size, phase percolation, etc.), the parallel kinetic

 

Seismic Wave Attenuation: Energy Dissipation in Crystalline Solids

255

mechanism that produces strain most rapidly will dominate the overall mechanical response, while the slowest serial step of that fastest parallel process will rate-limit the response. The major challenge in experimental geophysics is to understand the ranges of thermodynamic potentials giving rise to various mechanical responses in the laboratory time frame, and, also, to understand well the physics of their extrapolation: one must study in the laboratory the same physical process active in the geological environment for the data to be applicable (Paterson 1976). This point is as true for the phenomenon of seismic-frequency mechanical absorption as it is for large-strain creep phenomena. The maximum strains involved in wave attenuation are quite small, generally 10–6 for this grain size and temperature. It is logical to conclude, then, that a different physical mechanism begins to affect the absorption response with the increase in frequency/drop in temperature/increase in grain size conditions associated with the high-frequency “tail” on the spectrum. The “gentleness” of the deviation from the model suggests that it is the

Seismic Wave Attenuation: Energy Dissipation in Crystalline Solids

273

shear-inviscid grain boundary assumption in the model that is failing, that is, with the change in ƒ, T, d, the serial-kinetic process of grain boundary sliding begins to dissipate, in part, the mechanical energy. This point is the centerpiece of the Jackson et al. (2002) argument concerning their QG–1(d) data and its extrapolation in grain size.

Figure 12. Comparison of a 1200°C, polycrystalline olivine absorption master curve (created from 1200°C data from Jackson et al. 2002, Gribb and Cooper 1998, and all data from Bunton 2001; see text)—for an extrapolation to d = 1 mm—to the 1200°C torsional attenuation response of a forsterite single crystal that was first crept to steady state at 1600°C, 20 MPa (Gueguen et al. 1989). The tail in the master curve, associated with the absorption response(s) of viscid grain boundaries, maps into the response of the deformed single crystal. The correlation suggests low-angle (subgrain) boundaries play a significant—perhaps primary—role in the absorption of the deformed single crystal.

Because it acts in kinetic series with both diffusional and dislocation flow mechanisms—meaning that, under most thermodynamic conditions, it cannot be physically isolated—grain boundary sliding is difficult to characterize. At one extreme, the oft-cited model for grain boundary sliding-effected absorption treats grain boundaries as thin films of a finite width and a constant and finite Newtonian viscosity (Zener 1941). Loading produces sliding displacements on the boundaries, at a rate dependent on their unique viscosity, and elastic loading of triple junctions (hence the description of the physics as “elastically accommodated grain-boundary sliding,” e.g., Raj and Ashby 1971). From a mechanics-model perspective, one can imagine a Voigt/Kelvin Solid model placed in mechanical series with a spring (i.e., no plastic response is possible, only anelastic and elastic responses); the resultant model is one form of the “Standard Linear Solid,” and its dynamic response is an exponential-decay-based Debye peak. τ C in this case is given by (Nowick and Berry 1972, p. 437):

τ C = η b d / ξG ,

(21)

where η b is the boundary shear viscosity and G is the unrelaxed material shear modulus. The grain-size scaling for the peak location goes as τ C ∝ d, in that strain is characterized as the relative displacement across the boundary normalized by the grain size. An important ramification of the model is that the magnitude of the absorption is independent of grain size: a change in d solely causes a shift in frequency of the Debye peak with no change in its height.

274

Cooper

A nice example of the phenomenon is the absorption response of hot-pressed, polycrystalline Si3N4 ceramics (e.g., Mosher et al. 1976; Pezzotti et al. 1996). Si3N4 is one of the few materials where it is apparently all-but-impossible to form crystalline grain boundaries: in these materials, all grain boundaries have been replaced by ~1-nm thick films of amorphous silica or silicate (e.g., Clarke 1987). The chemistry (and thus the molecular structure and, with it, the viscosity) of these films is controlled by oxide additives. In general, the Newtonian viscosity of these thin films is >103–104 times smaller than that expected of diffusional processes in the Si3N4 proper. It is this extreme viscosity difference that allows experimental “access” to the boundary-sliding relaxation: to first order, the conditions envisioned by Zener (1941) (i.e., pure elastic accommodation of boundary sliding and little other time-dependent response) are met, quite clean Debye peaks are characterized and, from them, η b and its temperature sensitivity are determined. The Debye-peak grain-boundary absorption response matching the Zener (1941) model has been seen only rarely in metals. Kê’s (1947) work on high-purity aluminum is the classical example: the study revealed the single Debye peak whose magnitude correctly matched theoretical prediction and as well as the property of being independent of grain size; in addition, the τ C ∝ d was confirmed. Other careful studies comparing polycrystalline metals with single crystals (reviewed in Nowick and Berry (1972) and Lakes (1999)) periodically reveal a broad absorption peak of low magnitude in the polycrystalline material, one barely “competing” to be seen over the power-law background absorption. The interpretation of such a response is problematic. The simplest (albeit not unique) defensible interpretation is that the constant-boundaryviscosity assumption of the Zener (1941) model is, in general, inappropriate: actual grain boundary sliding (that is, in materials with crystalline grain boundaries, in contrast to the Si3N4 example cited earlier) involves chemical diffusion and the glide+climb motion of primary and secondary grain boundary dislocations (cf. Ashby 1972). In that grain boundary structure is a strong function of, for example, the relative orientation of grains, the loss of degeneracy of η b is easily contemplated. Of course, not even a broad grain-boundary-sliding peak is evident in the polycrystalline olivine data presented in Figures 11 and 12. This is perhaps not surprising: when the selection of thermodynamic and microstructural conditions is such that two (or more) serial-kinetic processes are vying for rate control, the impact on the measured kinetics is often subtle. Crossman and Ashby (1975) have modeled the plasticity behavior of polycrystals, specifically to gage the impact of the grain boundary response on the kinetics. The impact on bulk deformation kinetics of changes in thermodynamic conditions that cause the boundaries to evolve from inviscid to (uniformly) Newtonian-viscous to rigid is demonstrated to be very minor: at constant T, the effect is a ~10% increase in the stress providing a given strain rate, with the stress transition being played out over less than a half-order of magnitude in strain rate. Not minor, however, is the spatial distribution of strain (and thus of energy dissipation) within individual grains corresponding to the transition: for an overall Newtonian-viscous aggregate, the strain associated with viscous grain boundary sliding evolves from 15% of the total to 0%, but played out over two-to-three orders of magnitude in strain rate. Because of this serial-kinetic coupling of the diffusional based transient flow with the viscous response(s) of grain boundaries, simple application of superposition, e.g., as is done with parallel kinetic processes, is not physically sound. What can be said, then, of extrapolation of experimental data on grain-boundary effects in attenuation to the geological setting? For polycrystalline olivine, Jackson et al. (2002) have noted the complexities of physical behavior associated with the inviscid-to-

Seismic Wave Attenuation: Energy Dissipation in Crystalline Solids

275

viscous transformation of the grain boundaries, including the ineffectiveness of a straightforward application of superposition of the effects represented here by Equations (18) and (21), and so chose to apply the essentially empirical Schoeck et al. (1964) equation to their data, modified to include the apparent effect of grain size, i.e.,

Q −1 ∝ [ f −1d −1 exp(− Ea / RT )]α .

(22)

Their best-fit lines to those log Q–1 vs. log ƒ data at higher-frequency and lower temperature give a value for α of 0.27 and a value of Ea ~400 kJ mol–1. Jackson et al. (2002) continue by showing that a “vertical” translation (i.e., a shift in Q–1 with changes in grain size and/or temperature) of their data to grain sizes thought appropriate for the upper mantle produce Q–1(ƒ) results in rational agreement with values measured seismologically. Jackson et al. (2002) conclude, accordingly, that the “same grain-sizesensitive processes [active in the experiments] might be responsible for much of the observed seismic-wave attenuation.” Barring a better understanding of the physics of absorption of extended defects, like grain and phase boundaries, both Jackson et al.’s approach to data analysis and their conclusions must be considered reasonable. Discerning the physics of absorption processes of extended defects, and thus how absorption is affected by microstructure, is the present “cutting edge” of experimental research in attenuation/internal friction. Of interest in both geophysics and materials science is the specific impact of deformation-induced microstructure on the absorption behavior. Figure 12, thus, presents the provocative comparison of the polycrystalline olivine data with that for the attenuation response of a deformed forsterite single-crystal (Gueguen et al. 1989). In these experiments, the single crystal was first deformed to steady state at 1600oC and 20 MPa differential stress and, subsequently, a slice with minimum dimension 1 mm was tested in reciprocating shear at 1200°C. One sees clearly that the tail of the polycrystalline master curve “maps” into the single-crystal data. Further, there exists no high-angle grain boundaries to effect absorption (crystals not “pre-deformed” demonstrated a high-temperature experimental QG–1 ~ 10–2 that was essentially frequency independent). The result suggests strongly that low-angle (subgrain) boundaries in the single crystal, a result of the 20-MPa plastic deformation, are of primary importance in the overall attenuation response, that is, if the scale of these lowangle boundaries matches reasonably the 17–24 μm associated with the “tail” of the polycrystalline data. Such is the case: in a high-T, steady-state dislocation rheology, the single crystal must develop a polygonized microstructure, that is, there is developed a three-dimensional network of low-angle subgrain boundaries whose mean spacing (and spacing distribution) is set by the level of differential stress (and little-effected by T) (cf. Glover and Sellars 1973; Twiss 1977; Kohlstedt and Weathers 1980; Stone 1991). For forsterite at 1200°C, a 20 MPa differential stress creates a mean subgrain boundary spacing of ~20 μm. The physical implication, then, is that a small-amplitude, oscillatory variation in stress, one that is added to a nominally constant, and significantly greater differential stress that is effecting dislocation creep in a material, is “sampling” the deformationinduced microstructure, and particularly the network of subgrains. Such a perspective on the high-temperature background absorption behavior and its interrelation with the subgrain network is consistent fully with models attempting to formulate a microstructural foundation for the phenomenological mechanical (plasticity) equation of state. The equation-of-state approach (theory: e.g., Hart (1970); experiments in ionic solids: e.g., Lerner et al. (1979); Covey-Crump (1994)) suggests that a single, internal state-variable can be used to describe the strain-effected “hardness” (i.e., the microstructural resistance to flow) of a material. The model of Stone (1991) correlates this state variable to the statistical distribution of low-angle boundaries and how this

276

Cooper

distribution is affected by changes in stress and by accumulated strain: the steady-state distribution is primarily dependent on the applied stress, while the evolution of the distribution—effected by diffusion-based processes within the low-angle boundaries—is strain sensitive. Recent experiments on the flow of halite (Plookphol 2001), comparing/correlating the deformation kinetics and deformation-induced microstructures from both creep experiments and load-relaxation experiments, confirm many of Stone’s hypotheses. While this research correlating deformation-induced microstructure and “hardness” is young, one ramification is clear: with experimental care, the relative absorption effects of high-angle and low-angle boundaries (as well as of solid-state phase boundaries) can be deconvolved, physical models for these phenomena formulated and tested, and the results extrapolated. The possibilities of a more exacting and robust way of interpreting seismic Q–1 data, i.e., its use as a prospecting tool to understand structures and stresses (and perhaps strains) in active tectonic terranes, seems within the realm of reason. SUBGRAIN ABSORPTION AND THE ATTENUATION BAND

As evidence of the possibilities, consider the room-temperature attenuation spectra (tan δ ≡ QG–1), presented in Figure 13a, for various phases in the solder system indiumtin. These data are from Lakes’ laboratory, collected on his “broadband viscoelastic spectrometer” (Brodt et al. 1995); the breadth of frequencies explored is extraordinary. The data for γ-InSn4 (solid squares) and β-γ eutectic (solid triangles) data were published earlier (Lakes and Quackenbush 1996); the data for single-phase β-In3Sn (“×” symbol; between 10–1 and 103 Hz, these data are so dense that they appear as a solid line) represent a recently inaugurated collaboration with Lakes (McMillan et al., submitted). βIn3Sn is not an intermetallic compound but rather is an endothermic solid solution having the same body-centered tetragonal (BCT) crystal structure as pure In; as a consequence, β-In3Sn has a limited number of dislocation slip systems and the Peierls barriers to slip are high. Similarly, γ-InSn4 is also an endothermic solid solution with a hexagonal crystal structure, again having limited slip systems and a high Peierls barrier. β-γ eutectic is, thus, a two-phase (with different crystal structures) solid; in this case, γ grains are dispersed within a continuous, polycrystalline β phase. At frequencies in the range 10–3 to 1 Hz, one sees that the eutectic is significantly more absorbing than is either component phase individually: clearly, one cannot apply a simple rule-of-mixtures to model the data for these frequencies. Solid-state β–γ phase boundaries are perhaps so demonstrated to be more potent mechanical absorbers than β grain boundaries or subgrain boundaries in this frequency range; affirmation of this postulate requires a better understanding of absorption in the individual phases that constitute the eutectic. To better probe the physics of the broadband power-law absorption behavior, we have initiated a hierarchal study with single-phase β, using deformation and thermal processing to vary the dislocation (sub)microstructure. The baseline behavior—the ascast condition (labeled “tan δ; cast β In-Sn”)—is that presented in Figure 13a. The data display an attenuation plateau, m ~ 0, in the region 10–1–10 Hz. Beyond 10 Hz, the data trend back towards the m ~ 0.3 slope characteristic of the frequencies below 10–1 Hz; as such, a description of the data above 10–1 Hz as a “broad hump” is perhaps appropriate. (Note, too, absorption behavior of single-phase γ has a similar plateau for frequencies ≥10 Hz. The “tail” visible in the polycrystalline olivine data presented in Fig. 12 is also consistent with this spectral feature.) Figure 13b presents results of a parametric study in which the torsional absorption spectra (emphasizing the frequency range 10-1–103 Hz) of β specimens were measured as functions of various specimen preparations. These preparations were serial and cumulative, meaning that specimens were mechanically tested after each step of process-

Seismic Wave Attenuation: Energy Dissipation in Crystalline Solids

277

Figure 13. (a) Torsional tan δ (≡ QG–1) and normalized complex shear modulus (|G*|) as functions of frequency, and creep compliance (J(t)) as a function of time, for β-In3Sn (×), γ-InSn4 („) and the β–γ eutectic (solid triangles), all in the as-cast condition and at room temperature. (The tanδ data for β-In3Sn in the frequency range 10–1–103 Hz is so dense as to appear as a solid line.) One sees that the power-law absorption behavior holds for both singlephase β and γ materials at low frequency (m ≈ 0.3); both β and γ deviate from this behavior at increasing frequency (~10–1 Hz for β; ~1 Hz for γ). (b) Effect of processing conditions on the tanδ of single-phase β-In3Sn specimens. The processing followed a hierarchy of as-cast (“cast” in the figure); cast + pressed (high-rate, high-strain deformation; “pressed”); cast + pressed + annealed (“annealed”); cast + pressed + annealed + creep (“creep,” with the stress and strain noted). One sees a significant effect of the creep treatment: despite an increase in lattice dislocation density due to the tensile creep deformation, tan δ actually decreases. In addition, an absorption peak is created at ~102 Hz. Both results suggest an important perhaps dominant—role of low-angle boundaries in effecting the power-law absorption.

ing. Beyond the as-cast condition (spectrum marked “cast” in Fig. 13b, which is a replication of a portion of the spectrum shown in Fig. 13a), specimens were (i) plastically deformed at a high rate to greater than 60% compressive strain at 293 K (producing the absorption spectrum marked “pressed”), followed by (ii) annealing at 390 K (~95% of the

278

Cooper

melting temperature) for several hours (producing the spectrum marked “annealed”), followed by (iii) high-strain, plastic deformation in tension by dislocation creep (cf. Frost and Ashby 1982). Two different creep stresses were studied, 1.35 and 2.04 MPa. The absorption spectra are marked “creep;” the respective deformation stress and strain are noted as well. Optical microscopy of the specimens subsequent to various treatments revealed little effect of processing on grain shape or size: β grains remained nominally equiaxed with mean grain sizes ranging from ~500 μm for the “pressed” specimens to ~700 μm for the “annealed” specimens and, despite the large accumulated strain, creep subsequent to annealing had no statistically significant affect on the size or shape of the grains. In that torsional-specimen cross-sections were approximately 3×3 mm, the number of grains making up that cross-section was small. Torsional maximum strains in the absorption experiments were 10–6; despite the presence of lattice dislocations in the “pressed” and “creep” specimens, the materials were proven linearly viscoelastic at this strain. One sees clearly in the data (Fig. 13b) that, despite the creep deformation that must endow the specimens with a significantly higher density of mobile lattice dislocations than are remnant in an “annealed” specimen, the absorption of the crept specimens is significantly lower. Specifically, the spectra demonstrate a diminishment of the plateau/broad hump and demonstrate, too, the sharpening of a Debye peak near ƒ = 102 Hz. Because of the limited slip systems in the BCT-structure β, the tensile creep must result in a significantly limited distribution of subgrain boundary structures—a distribution in dynamic equilibrium with the applied tensile stress. The spectral peak at 102 Hz results from the absorption of a specific boundary structure (cf. Yan and Kê 1987), one obviously now more prevalent due to strain accumulation. Further, one sees that, for the “creep” specimens, absorption is apparently increased—actually, it is shifted to higher frequency—by deforming at a higher creep stress. The amount of this frequency shift is exactly that predicted by the change in τ (Eqn. 18), when d is considered to be the mean subgrain size, the magnitude of which is calculated from the mean-subgrainsize/differential-stress semi-empirical “calibration” of Twiss (1977) (and employing the measured, high-frequency shear modulus). An obvious question to ask, then, is whether the shallow peak near ƒ = 10–1 Hz in the absorption data for the deformed forsterite single crystal (Fig. 12) is related to a similar concentration of specific low-angle boundary structures. Gueguen et al. (1989) did not comment on this aspect of their data. Nevertheless, with limited slip systems, olivine, too, realizes a limitation on the structures of low-angle boundaries (e.g., Ricoult and Kohlstedt 1983; Bai and Kohlstedt 1992a). Finally, the diminishment of absorption with the addition of lattice dislocations to the material (Fig. 13b) should give pause to any generalized, qualitative speculation that “dislocation mechanisms” (i.e., the behavior of “free” lattice dislocations as opposed to the dislocation arrays constituting low-angle boundaries) are dominating a power-law, high-temperature-background absorption response. Inevitably, dislocation “vibration” with cyclic loading occurs, and must lead to nonlinearities in absorption (e.g., Lakes 1999, Ch.8). Yet under the thermodynamic (high temperature, low deviatoric stress), microstructural (polyphase, mineral (ionic) aggregates with each phase having a high Peierls barriers to dislocation glide) and frequency conditions of interest in geological settings, such mechanisms of absorption may prove of little consequence. IMPACT OF PARTIAL MELTING

The impact of partial melting on low-frequency mechanical absorption depends primarily on (i) the spatial distribution of the melt and crystalline phases, (ii) the relative

Seismic Wave Attenuation: Energy Dissipation in Crystalline Solids

279

volume fractions of these phases and (iii) the practical rheological contrast of these phases. Early geophysical analyses to correlate materials properties with their capability to absorb shear waves concentrated on the spatial distribution of a liquid phase at the scale of the grain size of the crystalline residuum (e.g., Walsh 1969; Birch 1970; Mavko 1980). For example, it is easy to comprehend that a partial-melt microstructure in which all solid-state grain and phase boundaries are replaced by a homogeneous, low-viscosity liquid will be far more absorbing of shear waves (at a fixed frequency related to the viscosity of the liquid films—cf. the Si3N4 example cited earlier) than a system where grain and phase boundaries remain melt-free. Waff (e.g., Waff and Bulau 1979; Waff 1980; Waff and Faul 1992) inaugurated the petrologic study of partial-melt morphology, which followed in philosophy the microstructure topology work of Smith (1952). Waff defined “textural equilibrium” as the minimization of solid-solid (grain boundary or solid-state phase boundary) and solid-liquid interfacial energy; in olivine-basalt partial melts, under hydrostatic conditions and for a small volumetric melt fraction, textural equilibrium sees the liquid phase confined to three-grain (“triple”) junctions and fourgrain corners, forming a 3-D interconnected melt network, with melt-free olivine grain boundaries. That the grain boundaries in these materials were indeed melt free was confirmed to a resolution of 0.5 nm using high-resolution transmission electron microscopy (HRTEM; e.g., Vaughan et al. 1982; Cooper and Kohlstedt 1982), a result confirmed by high-resolution transmission analytical electron microscopy (HRAEM; Kohlstedt 1990). The result remains controversial: for example, de Kloe et al. (2000) argue from HRTEM/AEM evidence that a deviatoric stress (≥100 MPa) changes this textural equilibrium so as to produce nm-scale melt films that replace grain boundaries. The chapter in this volume by David Kohlstedt, examining the rheology of partial melts and the physics of melt segregation, presents the breadth of microstructural evidence and associated arguments. Nevertheless, in drained mafic-to-ultramafic material at low melt fraction, the vast majority of the liquid phase is confined to triple junctions and four-grain corners. The impact of grain-scale, partial-melt textural equilibrium on attenuation, then, involves three possible effects that are perhaps distinguishable experimentally: (i) longrange (i.e., beyond the grain scale) melt migration based on gradients in hydrostatic pressure (a bulk-modulus-mode absorption); (ii) melt “squirt” (Mavko and Nur 1975), a shear effect that adjusts the volume of adjacent triple junctions at the grain scale; and (iii) any shear effect that might arise from melt films replacing grain boundaries and/or solidstate phase boundaries. Long-range melt migration is best contemplated following a D’Arcy flow model in which the compaction of the crystalline residuum imparts momentum to the interconnected liquid phase; in the geophysics literature, the seminal papers of McKenzie (1984), Richter and McKenzie (1984) and Scott and Stevenson (1986) present these ideas, now known as “compaction theory.” The relative velocities of the crystalline residuum and the liquid are linearly proportional to the inverse gradient of the pore fluid pressure, with the constant of proportionality being the permeability of the residuum, i.e., u − U = − ⎡⎣ k (φ , d ) / η f ⎤⎦ ∇Π ,

(23)

where

(

)

∇Π = η∇ 2U + (ζ + η / 3) ∇ ( ∇ ⋅ U ) − (1 − φ ) g Δρ − H γ sl , φ ∇φ .

(24)

In Equation (23), u and U are the velocities of the melt and crystalline residuum, respectively, k is the permeability (which is a function of the volumetric melt fraction, φ, and the grain size), η f is the shear viscosity of the liquid and ∇Π is the gradient in pore

280

Cooper

fluid pressure; in Equation (24), η is the shear viscosity and ζ the bulk viscosity of the partial-melt aggregate, respectively, g is the acceleration of gravity, Δρ is the density difference between liquid and crystalline phases and γ sl is the interfacial energy (“surface tension”) of the crystal (solid)-liquid interface. The first two terms in Equation (24) represent pressure gradients affected by shear and dilatational deformation, respectively, the third term is the buoyancy effect and the last term (Stevenson 1986; Cooper 1990) is the restraining effect of capillarity. In the geophysical context, gravity drives the liquid phase upwards while the residuum compacts downward; capillarity restrains the expulsion of all of the melt. From a chemical kinetics perspective, buoyancy-driven melt migration is obviously a serial-kinetic process based on a geometrical constraint: liquid is not expelled unless the residuum compacts. Parametric analysis of the equations reveals a characteristic “compaction length,” δ c:

δ c = [k (ζ + 4η / 3) / ηf ]1/ 2 ;

(25)

δ c is the instantaneous length over which the compaction is rate limited solely by the

capabilities of the residuum to deform (i.e., the actual flow of the liquid phase dissipates none of the energy in compaction). Beyond δ c, energy dissipation is convolved between the deformation of the residuum and shear flow of the viscous liquid, although at scales large compared to δ c, the shear flow of the liquid should be rate-limiting.

Bulk-modulus anelasticity for a partial melt arises due to the restorative force of capillarity: a hydrostatic compressional force promotes compaction; capillarity (the Laplace-Young relationship, e.g., that defines the correlation of the pressure in a bubble to its surface tension) promotes swelling upon removal of the applied force. Flexural creep experiments, which optimize the melt-migration driving force while minimizing the driving force for shear creep of the aggregate, combined with both microstructural analyses and strain-relaxation measurements, proved the operation of the long-distance migration mechanism and its anelastic nature (Cooper 1990; Gribb et al. 1994). The seismic-frequency absorption response for this mechanism was measured directly for partially molten, Newtonian-viscous silicate partial melts in reciprocating flexure ((ortho)enstatite glass-ceramic, Gribb and Cooper (1995)) and reciprocating compression (Co-Mg olivine-“basalt”, Gribb et al. (1994)). In both these cases of Young’s-modulusmode loading, the effective δ c was significantly smaller than the specimen size: flow of the liquid phase rate-limited the anelastic response. The absorption spectra revealed the power-law behavior, with a very-low m ~ 0.10–0.14 (see Fig. 6 for the glass-ceramic data). As noted earlier, a single, non-exponential-decay mechanism related to the D’Arcy-flow (diffusion)-effected migration of the flexural-specimen’s neutral axis (or to the growth of the compaction layer—from the specimen’s free sides to the specimen center—in compression specimens) accurately models the absorption response for this long-range melt migration. Given the thermodynamics of capillary forces, the long-distance-migration mechanism is not expected to be operative in pure-shear loading. As a consequence, shear experiments can isolate the other two, grain-scale partial-melt absorption effects listed above. Gribb and Cooper (2000) compared the shear absorption responses of finegrained (d ~ 3 μm) olivine aggregates with and without the addition of 5 wt % (6 vol %, i.e., φ ~ 0.06) of a texturally equilibrated, K+-doped basanite melt (the melt composition was selected to be saturated with respect to olivine as well as to have a low-enough viscosity to provide a δ c that was significantly greater than the specimen size). Experiments were done in the regime of (ƒ,d,T) characteristic of the shear-inviscid boundary assumption being applicable to the melt-free specimens. The 1250oC result, characteristic of all the experiments, is presented in Figure 14a: the partial melt specimens showed essentially the same power-law absorption spectra as do the melt-free

Seismic Wave Attenuation: Energy Dissipation in Crystalline Solids

281

Figure 14. Effects of grain scale partial melting—at low melt fraction and the textural equilibrium corresponding to nominally hydrostatic pressure (melt-free grain boundaries)—on the shear attenuation response of olivine. (a) Sub-resonant behavior. At the high-temperature and/or low-frequency conditions where the inviscid grain boundary assumption holds, partial melting affects the absorption spectrum solely by its shifting to slightly higher frequency. The magnitude of the frequency shift is predicted by the effect of partial melting on the aggregate η ss, following the model of Cooper et al. (1989). There is no apparent disaggregation of the material—with consequent dramatic drop of shear modulus—because of the presence of melt. [Used by permission of American Geophysical Union, from Gribb and Cooper (2000), Geophysical Research Letters, Vol. 27, Fig. 3, p. 2343.] (b) Resonant behavior at ƒ ~2 Hz. Two Debye peaks are revealed: the one at high temperature may be related to the grain-scale “melt squirt” phenomenon envisioned by Mavko and Nur (1975); alternatively, it may represent a solid-state grainboundary absorption. [Used by permission of Columbia University Press, from Xu et al. (2002). In MARGINS Theoretical and Experimental Earth Science, edited by G. Karner, Fig. 10.3, in press.]

specimens (QG–1 ∝ ƒ –0.4), with the partial melt spectra shifted to a higher frequency that corresponds to an decrease in η ss of a modest factor of 6. In the figure, a shift of the Andrade-Model inversion for the melt-free material, corresponding to this decrease in η ss for the partial melt, predicts beautifully the partial melt attenuation. The important interpretation is that relative to point (iii) above: there is no melt-effected plummeting of viscosity or shear modulus, nor a dramatic increase in shear attenuation, as might be anticipated if all grain boundaries were replaced by films of a low-viscosity liquid. Further, the thermal sensitivity (activation energy) of both creep and attenuation responses were identical between partial-melt and melt-free materials, clearly suggesting that the same physical process rate-limits both creep and attenuation in both materials.

282

Cooper

The results (i.e., the magnitude of decrease in η ss and the consequent effect on the attenuation spectra) are consistent fully with the model for melt-enhanced diffusional creep in which diffusion through grain boundaries is rate limiting and the triple-junction melt network, representative of textural equilibrium, provides a partial short-circuit for diffusion (Cooper et al. 1989; see the chapter in this volume by Kohlstedt). Shear resonance (torsion pendulum) experiments (Xu et al. 2002) performed on finegrained (d ~ 20 μm) olivine-basalt aggregates (1 ≤ ƒ[Hz] ≤ 10; 22 ≤ T[oC] ≤ 1350; 0.004 ≤ φ ≤ 0.13) reveal two distinct Debye peaks, one centered on T ≈ 875oC and the other on T ≈ 1175oC (the data for φ = 0.06 is presented in Fig. 14b). The high-T peak has a magnitude that is weakly sensitive to φ : QG–1 ~0.3 for φ = 0.13; QG–1 ~0.1 for φ = 0.004. The low-temperature peak is associated with the glass transition of the quenched basalt; the high-temperature peak, however, speaks either of a melt-squirt effect and/or of a grain-boundary effect in shear absorption. Melt squirt is defined as an exponential-decay absorption mechanism with a characteristic time, τ ms, given by (cf. Mavko 1980):

ηf = (τ ms K f / 40) ( R / d ) 2 ,

(26)

where Kf is the bulk modulus of the liquid and R is a measure of the cross-section of the triple-junction “tube.” Xu et al. (2002) calculate that, for the characteristic τ C of their experiments, the η f is too high, by an order of magnitude, than that of a bulk, olivinesaturated liquid basalt; as a consequence, they interpret the high-temperature peak to be one associated with the Zener grain-boundary absorption (i.e., Eqn. 21). Their argument continues by demonstrating that an extrapolation of their data to φ = 0 matches well the “plateau” result of Bunton (2001) for melt-free olivine of similar grain size. It seems premature, however, to completely rule out a melt-squirt contribution in this absorption peak. For example, the viscosity of a bulk liquid may not accurately reflect the shear resistance of a silicate melt within a ≤1 μm “diameter” triple junction tube, particularly considering that the melt flow involves, too, stress-effected chemical reaction with the tube (cf. Riley and Kohlstedt 1991). Further, the significant difference in the thermal sensitivity for their background absorption (Ea ~600–650 kJ mol–1, which is related to grain boundary diffusion and, most likely, to grain boundary viscosity as well) compared to that for the measured τ C of the Debye peak (Ea ~300–350 kJ mol–1) suggests a distinctly different physical process controlling the latter phenomenon. A melt-induced dissipation peak superposed on the power-law absorption response of melt-free olivine has also been observed recently in sub-resonant torsion experiments by Jackson et al. (2001). A critical factor not yet addressed in any attenuation experiments on partial melts is the role played in absorption by incipient melt segregation beyond the grain scale. Geochemical evidence strongly indicates that melt flow at the grain scale must be very limited in extent, that is, upon formation of melt it must be focused into larger channels so as to avoid continuous reaction with the crystalline residuum (e.g., Spiegelman and Kenyon 1992). Arguments based on the thermodynamics of irreversible processes suggests that a fractal drainage system should prove effective to this end (e.g., Hart 1993). Consistent with the fractal hypothesis, the incipient process of beyond-grain-scale phase separation of melt and crystalline residuum, driven by deformation and affected in scale by the relative values of (bulk-specimen) strain rate and δ c, has recently been demonstrated experimentally (Holtzman et al. 2001). Correlation of this scaling with the seismic-frequency absorption response for the material overall is an experimental challenge with obvious important implications.

Seismic Wave Attenuation: Energy Dissipation in Crystalline Solids

283

IMPACTS OF DEFECT CHEMISTRY (INCLUDING WATER) AND THE STRUCTURE(S) OF INTERFACES

The rates of chemical diffusion and of temperature-dependent rheology of crystalline solids depend directly on the concentrations of point defects on the crystal lattice. These point defects include vacancies (missing ions or atoms), interstitials (ions/atoms placed in the structure in interstices not occupied in the “perfect” crystal structure) and substitutionals (e.g., as occurs in a solid solution); they additionally include bonded complexes of these single defects as well as the jogs and kinks on lattice dislocations. Made thermodynamically stable by the configurational entropy of their distribution on the lattice, the concentrations of point defects are set by the temperature, pressure (in the limit, they can perhaps be set solely by temperature and pressure—these are the intrinsic point defects), deviatoric stress, and by the various chemical potentials defining chemical equilibrium in the system. Indeed, the concentrations of point defects are the physical manifestation of chemical activities within a crystal. In minerals—because the bonding is primarily ionic and, thus, the energy gap between the valence and conduction bands is so large—point defects carry a charge; as such, equilibrium requires that the product of concentration and valence of the positively charged defects equals that of the negatively charged defects; this condition is part of the overall charge neutrality requirement for the crystal. The large band gap also insures that, in the geologic setting, point defect concentrations are affected significantly by changes in the activities (fugacities) defining the equilibrium; this situation is defined as the extrinsic condition. In that most minerals contain transition-metal cations, even in very small concentrations, additional important point defects include conduction electrons and electron holes, which are (for example) transition-metal cations having either a too-small valence or a too-large valence, respectively, relative to that valence required for neutrality on the cation site of interest. The concentrations of conduction electrons and electron holes, and therefore (because of the charge-neutrality constraint) the concentrations of most of the other point defects on the lattice, are consequently affected significantly by the oxygen fugacity. (The best pedagogical treatise on point-defect thermodynamics in ionic materials and its application to dynamic processes is the text by Schmalzried (1995).) The effect of point defect concentrations on mechanical absorption is fourfold (cf. Karato 1995): (i) the defects can perhaps affect the bond strength in the material, and so affect the modulus; (ii) at very high concentrations, the defects can bond and this polar “complex” can respond to stress in a similar manner to that producing the Snoek relaxation described earlier; (iii) the defects affect the rates of all dynamic processes including chemical diffusion and the climb and glide of lattice dislocations and of primary and secondary grain boundary (phase boundary) dislocations; and (iv) the defects affect the structure of grain boundaries and of solid-state phase boundaries, which, in ionic solids, also includes an electrostatic space charge penetrating into the grains on either side of the boundary. Because, in general (though not including substitutional solid solutions of widely ranging concentration), the total concentration of point defects is usually very small, effect (i) can be considered of little consequence. Similarly, the formation of defect complexes is prevalent at lower temperatures and the frequencies anticipated for their mechanical absorption are far higher than those of interest seismologically; effect (ii) can also be set aside. Effects (iii) and (iv), however, are significant to this discussion. The dynamic effect of point defect concentration (point (iii) above) is often probed by measuring the impact of various chemical fugacities on, for example, viscosity (or, in non-linear systems, effective viscosity). What is discovered experimentally (and supported by point-defect thermodynamics theory) is a power-law dependence, that is,

284

Cooper

viscosity is characterized as: a b η ∝ f ox1 f ox2 LfOc exp( Ea / RT) , 2

(27)

where f j is fugacity of species j (ox1 and ox2, etc., representing different crystalline oxides that are components of the phase(s) whose viscosity is being probed). The values of power-law exponents a, b, and c (± the value of Ea) determined in experiments are compared to theoretical models to discern the specific point defects affecting/effecting the physical behavior. The careful work of Bai et al. (1991) probing the impacts of oxygen fugacity and orthopyroxene (or magnesiowüstite) activity on the dislocation creep rate of single-crystal olivine is a fine example of the application of point defect thermodynamics to mechanical response. It has long been known that water fugacity ( f H 2O ) has a dramatic effect on the viscosity of silicates (both crystalline and amorphous), decreasing it dramatically. Griggs (1967) described the phenomenon as “hydrolytic weakening;” the experimental rock physics community has been exploring the phenomenology of this chemical/mechanical effect ever since (the chapters by Jan Tullis and by Greg Hirth in this volume detail the current understanding). The phenomenology suggests clearly that hydrolytic weakening is best contemplated from the perspective of (extrinsic) point defect thermodynamics (i.e., relationships similar in form to Equation (27) can be defined for f H 2O -affected viscosity; e.g., Hobbs (1981)). Identifying the specific point defect(s) responsible for hydrolytic weakening (which could vary for different silicate phases) remains difficult and controversial; the experimental work of Kohlstedt and co-workers, for example, on the reaction of water with various polymorphs of (Mg,Fe)2SiO4 illustrates well the confluence of experiment with point-defect theory (e.g., Mackwell and Kohlstedt 1990; Bai and Kohlstedt 1992b; Kohlstedt et al. 1996). These studies additionally suggest that the primary source of water in the mantle is not that incorporated in the (“perfect”) structure of hydrous minerals, but rather as extrinsic point defects in the “anhydrous” minerals. To zero-th order, then, the impact of chemical environment (via its effect on pointdefect concentrations—including water-related extrinsic defects) on attenuation is straightforward: impacts on viscosity cause changes in the various τ C’s with consequent frequency shifts in the Q–1 spectra. For example, in that recent grain boundary diffusion creep studies on polycrystalline olivine demonstrate that η ss ∝ f H 2O -1 (Mei and Kohlstedt 2000; cf. Karato et al. 1986), the effect on the absorption model presented in Equations (18) and (20) (Fig. 9) is easily calculated. There are, however, more profound issues that, while not yet explored directly experimentally, bear contemplation with regard to the impact of point defects, and particularly water-related defects, on the mechanical absorption of mantle material. In partial melting, for example, partitioning of water between the crystalline residuum and the silicate melt functions both to increase the solid-state viscosity of the olivine and lower that of the melt; calculations of the partitioning, integrated with the understanding of its effects on rheology (Karato 1986; Hirth and Kohlstedt 1996), affect the understanding of upper-mantle dynamics, particularly melt segregation. As for attenuation, then, the zeroth-order effects of crystalline point defects are convolved with the partial melt effects noted earlier. In that the low-frequency, power-law absorption behavior seems intimately related to the density, distribution and structure of interfaces (grain boundaries, solid-state phase boundaries, subgrain boundaries, ± solid-liquid phase boundaries) the interaction of lattice point defects with the structural elements of these interfaces is of critical importance to a quantitative understanding of attenuation. In the solid state, the robust

Seismic Wave Attenuation: Energy Dissipation in Crystalline Solids

285

understanding of grain/phase boundaries to have a 2-D crystallographic structure based on dislocations (Sutton and Balluffi 1995) means that the jogs and kinks on these boundary dislocations (as well as those on the “free” lattice dislocations) contribute to the point-defect equilibrium of the system. Thus the kinetics of stress-induced motion of the extended defects is affected to first order by the extrinsic chemical condition. The interaction (in the case of ionic solids) of charged jogs and/or kinks with charged lattice point defects can either inhibit or enhance the motion of the extended defect—as well as the temperature sensitivity of that motion (cf. Maier 1994; Jamnik and Raj 1996; Chiang et al. 1997, p.155). (The case of rate inhibition is fully analogous to the elastic interactions between solute atoms and dislocations/grain boundaries in metals that results in “solid-solution strengthening” (e.g., Courtney 1990, p.173).) The phenomenological experimental record of the impact of f H 2O on the kinetics of dislocation creep and on fabric development of olivine is an important example. An increase in f H 2O , while lowering the effective viscosity overall, is known to effect relative changes in the resistance to dislocation motion of different slip systems in olivine (e.g., Mackwell et al. 1985); an increase in f H 2O has also been correlated with an increase the mobility of grain boundaries (e.g., Jung and Karato 2001a). The combination of the two f H 2O effects, in the presence of a constant deviatoric stress, has been demonstrated to result in distinctly different fabric (i.e., lattice preferred orientation) for the deformed material compared to that seen for nominally “dry” conditions (Jung and Karato 2001b). The impact on seismic responses are (i) to change significantly the wave velocity anisotropy, that is, its directionality relative to the principal stress directions, and (ii) to decrease the variety of grain boundary structures in the deformed material (a geometrical requirement consistent with the development of fabric), and so affect the absorption spectrum (cf. the case for the β-In3Sn creep specimens presented in Fig. 13b, where one aspect of fabric—spatial distribution of low-angle boundary structures—has a notable, demonstrable effect on absorption). The fabric effect presented above is based on the kinetic response of continued, accumulated strain in an polycrystalline aggregate. But, as noted earlier, the attenuation response is that associated with a very small stress pulse that adds to the overall stress potential giving rise to the large-scale deformation. Both the crystallographic structure and the electrostatic structure of the interfaces dictates the response: the stress-pulseinduced climb and glide of primary and secondary grain (phase)-boundary dislocations are subject both to their specific crystallography and to the electrostatic interactions with lattice point defects. The electrostatic and crystallographic structures are coupled to the extrinsic point-defect chemistry. A dramatic, and quantitatively characterized, example of the boundary-structure/extrinsic-defect relationship is the case of “twist” boundary (boundary normal and twist axis both [001]) structure in pure (BCC) α-iron and α-iron doped with 0.18 at % gold (Sickafus and Sass 1985). TEM and ion-backscattering spectroscopy proved that the impurity gold atoms preferentially segregate to the grain boundary, and the segregation results in a two-dimensional structural reaction. In pure αiron, the boundary consists of an orthogonal array of [1 10] and [110] screw dislocations; with the addition of the impurity, the boundary transforms to an orthogonal array of [010] and [100] screw dislocations. It is not a great stretch to believe that similar transformations occur in ionic solids; certainly, impurity segregation to boundaries has been well characterized in variety of oxides and other ionic compounds (e.g., Chiang et al. 1997, p.156). The impact of these chemical-segregation effects and their possible influence on boundary structure and viscosity, etc. are now being contemplated by the experimental community. For example, very small concentrations of Ca2+ in olivine apparently effect a significant change in the thermal sensitivity of grain boundary diffusional creep (Ea in the

286

Cooper

range 300–400 kJ mol–1 for aggregates prepared from pulverized San Carlos peridot, with CaO content of ~0.05 wt %; Ea in the range 600–700 kJ mol–1 for aggregates prepared either from a dunite with a CaO content of ~0.3-0.6 wt % or for San Carlos specimens prepared with very small amounts of CaO-bearing basaltic melt). (There are copious examples of similar behavior in the ceramic materials literature, e.g., Cho et al. (1997).) Similarly, the η ss ∝ d 2 result for diffusional creep in fine-grained olivine at small deviatoric stress (noted in Section 7) seems consistent with a threshold effect related to the electrostatic structure(s) of the grain boundaries. And the exact impact of waterrelated extrinsic defects on boundary-effected absorption have yet to be characterized. Because the stresses (and strains) involved in seismic wave attenuation are so small—and the extrapolation goals for the experimental research so grand!—these chemical effects may prove of first-order importance and so require close scrutiny in future experiments. FINAL COMMENTS

This chapter has sought specifically to be a primer on linear viscoelasticity at high temperature and low frequency, conditions characterizing wave attenuation in geological systems. The emphasis placed on appreciating viscoelastic models and their impact on thinking with regard to attenuation is to allow the reader a base for access to the literature. The geophysical problem is a difficult one: the power-law absorption spectrum is fairly non-descript; and, despite one quantitative characterization (the shear-inviscidboundary, diffusional creep model giving rise to Fig. 9) that can forward-predict the absorption-band behavior for a finite window of thermodynamic and microstructural conditions, the overall absorption behavior of the terrestrial planets undoubtedly involves contributions of a number of serial-kinetic mechanisms (and maybe some parallel kinetic mechanisms as well), which are difficult, indeed, to isolate experimentally. Progress in characterizing attenuation response(s) quantitatively and with predictive capability requires advancement on a number of fronts. First, there is significant need for an improved understanding of the physical processes involved in crystalline plasticity, particularly in the areas of statistical studies of dislocations and dislocation arrays, e.g., correlating phenomenological mechanical-equation-of-state approaches to plasticity with the more classical, steady-state creep and microstructural studies, and in the application of irreversible-process thermodynamics to plasticity problems in polyphase crystalline aggregates. Second, in dynamic experiments, the role of threshold phenomena will need to be studied directly. Critical for rock and mineral studies will be the multiple impacts of various chemical potentials—including, particularly, that of H2O—on the distributions of lattice point defects and on the structure and related viscosity of grain and phase boundaries. Nevertheless, given recent experimental and theoretical advances in the mineral physics of mechanical absorption, along with the inspiration of various geodynamic models, there is reason for optimism that an experimentally calibrated, but physically based “key” can be developed that will correlate the magnitude and frequencysensitivity of seismic wave absorption with the stresses and accumulated strains in tectonic terranes. ACKNOWLEDGMENTS

I have been blessed with a number of graduate students—Tye Gribb, Joe Bunton, Jeff Lee and Sam Zhang—and one postdoc—Doug Green—who have pursued with me various aspects of the attenuation problem in polycrystals and partial melts; all were outstanding collaborators and I am pleased to acknowledge them for their efforts and continued friendship. The experimental wave attenuation community in geophysics is a fairly small and collegial group, from whom I have learned a great deal; in particular, periodic discussions with Ian Jackson, Shun Karato, David Kohlstedt, Brian Bonner and

Seismic Wave Attenuation: Energy Dissipation in Crystalline Solids

287

Ivan Getting have proven distinctly valuable. Likewise, in the materials and mechanics community, active collaborations with Rod Lakes and Donald Stone, and periodic discussions with Rishi Raj, continue to sharpen my perspective. The attenuation research pursued in my group at Wisconsin has been funded, in part, by the National Science Foundation Division of Earth Sciences Program in Geophysics; that support is gratefully acknowledged. REFERENCES Anderson DL, Given JW (1982) Absorption band Q model for the Earth. J Geophys Res 87:3893-3904 Anderson DL, Minster JB (1979) The frequency dependence of Q in the Earth and implications for mantle rheology and Chandler wobble. Geophys J R Astron Soc 58:431-440 Andrade, ENDaC (1910) On the viscous flow in metals, and allied phenomena. Proc R Soc London A84: 1-12 Ashby MF (1972) Boundary defects and atomistic aspects of boundary sliding and diffusional creep. Surf Sci 31:498-542 Bai Q, Kohlstedt DL (1992a) High-temperature creep of olivine single crystals, 2. Dislocation structures. Tectonophysics 206:1-29 Bai Q, Kohlstedt DL (1992b) Substantial hydrogen solubility in olivine and implications for water storage in the mantle. Science 357:672-674 Bai Q, Mackwell SJ, Kohlstedt DL (1991) High-temperature creep of olivine single crystals, 1. Mechanical results for buffered samples. J Geophys Res 96:2441-2463 Berckhemer H, Kampfmann W, Aulbach E, Schmeling H (1982), Shear modulus and Q of forsterite and dunite near partial melting from forced-oscillation experiments. Phys Earth Planet Inter 29:30-41 Birch F (1970) Interpretations of the low velocity zone. Phys Earth Planet Inter 3:178-181 Brady JB (1995) Diffusion data for silicate minerals, glasses and liquids. In Mineral Physics and Crystallography: A Handbook of Physical Constants. Ahrens TJ (ed) American Geophysical Union, Washington, p 269-290 Brodt M, Cook LS, Lakes RS (1995) Apparatus for determining the properties of materials over ten decades of frequency and time: Refinements. Rev Sci Instrum 66:5292-5297 Bunton JH (2001) The impact of grain size on the shear creep and attenuation behavior of polycrystalline olivine. MS thesis, University of Wisconsin–Madison, Madison, Wisconsin Chan WW, Sacks IS, R. Morrow R (1989) Anelasticity of the Iceland Plateau from surface wave analysis. J Geophys Res 94:5675-5688 Chiang Y-M, Birnie D III, Kingery WD (1997) Physical Ceramics: Principles for Ceramic Science and Engineering. Wiley, New York Cho JH, Harmer MP, Chan HM, Rickman JM, Thompson AM (1997) Effect of yttrium and lanthanum on the tensile creep behavior of aluminum oxide. J Am Ceram Soc 80:1013-1017 Clarke DR (1987) On the equilibrium thickness of intergranular glass phases in ceramic materials. J Am Ceram Soc 70:15-22 Coble RL (1963) A model for boundary diffusion controlled creep in polycrystalline materials. J Appl Phys 34:1679-1682 Cook LS, Lakes RS (1995) Viscoelastic spectra of Cd0.67Mg0.33 in torsion and bending. Metall Mater Trans A 26A:2035-2039 Cooper RF (1990) Differential stress-induced melt migration: An experimental approach. J Geophys Res 95:6979-6992 Cooper RF, Kohlstedt DL (1982) Interfacial energies in the olivine-basalt system. Adv Earth Planet Sci 12:217-228 Cooper RF, Kohlstedt DL, Chyung K (1989) Solution-precipitation enhanced diffusional creep in solidliquid aggregates which display a non-zero dihedral angle. Acta Metall 37:1759-1771 Courtney TH (1990) Mechanical Behavior of Materials. McGraw-Hill, New York Covey-Crump SJ (1994) The application of Hart’s state variable description of inelastic deformation to Carrara marble at T < 450oC. J Geophys Res 99:19793-19808 Crossman FW, Ashby MF (1975) The non-uniform flow of polycrystals by grain boundary sliding accommodated by power-law creep. Acta Metall 23:425-440 D’Anna G, Benoit W, Vinokur VM (1997) Internal friction and dislocation collective pinning in disordered quenched solid solutions. J Appl Phys 82:5983-5990 de Kloe R, Drury MR, van Roermund HLM (2000) Evidence for stable grain boundary melt films in experimentally deformed olivine-orthopyroxene rocks. Phys Chem Minerals 27:480-494 Dooris A, Lakes RS, Myers B, Stephens N (1999) High damping indium-tin alloys. Mech Time-Depend Mater 3:305-318

288

Cooper

Findley WN, Lai JS, Onaran K (1976) Creep and Relaxation of Nonlinear Viscoelastic Materials. NorthHolland, Amsterdam (corrected and republished: Dover Publications, New York, 1989) Frost HJ, Ashby MF (1982) Deformation Mechanism Maps: The Plasticity and Creep of Metals and Ceramics. Pergamon Press, Oxford, UK Glover G, Sellars CM (1973) Recovery and recrystallization during high-temperature deformation of αiron. Metall Trans 4:765-775 Green DH, Cooper RF (1993) Dilatational anelasticity in partial melts: Viscosity, attenuation and velocity dispersion. J Geophys Res 98:19807-19817 Green DH, Cooper RF, Zhang S. (1990) Attenuation spectra of olivine/basalt partial melts: Transformation of Newtonian creep response. Geophys Res Lett 17:2097-2100 Gribb TT (1992) Low-frequency attenuation in microstructurally equilibrated silicate partial melts. MS thesis, University of Wisconsin–Madison, Madison, Wisconsin Gribb TT, Cooper RF (1995) Anelastic behavior of silicate glass-ceramics and partial melts: Migration of the amorphous phase. In Plastic Deformation of Ceramics. Bradt RC, Brookes CA, Routbort JL (eds) Plenum Press, New York, p 87-97 Gribb TT, Cooper RF (1998) Low-frequency shear attenuation in polycrystalline olivine: Grain boundary diffusion and the physical significance of the Andrade model for viscoelastic rheology. J Geophys Res 103:27267-27279 Gribb TT, Cooper RF (2000) The effect of an equilibrated melt phase on the shear creep and attenuation behavior of polycrystalline olivine. Geophys Res Lett 27:2341-2344 Gribb TT, Zhang S, Cooper RF (1994) Melt migration and related attenuation in equilibrated partial melts. In Magmatic Systems. Ryan MP (ed) Academic Press, San Diego, p19-36 Griggs DT (1967) Hydrolytic weakening of quartz and other silicates. Geophys J Roy Astron Soc 14:19-31 Gueguen Y, Darot M, Mazot P, Woirgard J (1989) Q–1 of forsterite single crystals. Phys Earth Planet Inter 55:254-258 Hart EW (1970) A phenomenological theory for plastic deformation of polycrystalline metals. Acta Metall 21:295-307 Hart SR (1993) Equilibrium during mantle melting: A fractal tree model. Proc Natl Acad Sci USA 90:11914-11918 Herring C (1950) Diffusional viscosity of a polycrystalline solid. J Appl Phys 21:437-445 Hirth G, Kohlstedt DL (1995) Experimental constraints on the dynamics of the partially molten upper mantle: Deformation in the diffusion creep regime. J Geophys Res 100:1981-2001 Hirth G, Kohlstedt DL (1996) Water in the oceanic upper mantle: Implications for rheology, melt extraction and the evolution of the lithosphere. Earth Planet Sci Lett 144:93-108 Hobbs BE (1981) The influence of metamorphic environment upon the deformation of minerals. Tectonophysics 78:335-383 Holtzman B, Zimmerman ME, Kohlstedt DL, Phipps Morgan J (2001) Interactions of deformation and fluid migration I: Melt segregation in the viscous regime. EOS Trans, Am Geophys Union 82:F1107 Jackson I (1993) Progress in the experimental study of seismic wave attenuation. Ann Rev Earth Planet Sci 21:375-406 Jackson I (2000) Laboratory measurement of seismic wave dispersion and attenuation: Recent progress. In Earth’s Deep Interior: Mineral Physics and Tomography from the Atomic to the Global Scale. Karato S-i, Forte AM, Liebermann RC, Masters G, Stixrude L (eds) American Geophysical Union, Washington, p 161-179 Jackson I, Faul UF, FitzGerald JD (2001) Laboratory measurements of seismic wave attenuation in uppermantle materials: the effect of partial melting. EOS Trans, Am Geophys Union 82:F1163 Jackson I, FitzGerald JD, Faul UF, Tan BH (2002) Grain-size sensitive seismic-wave attenuation in polycrystalline olivine. J Geophys Res (in press) Jackson I, Paterson MS, FitzGerald JD (1992) Seismic wave dispersion and attenuation in Åheim dunite: An experimental study. Geophys J Intl 108:517-534 Jamnik J, Raj R (1996) Space-charge-controlled diffusional creep: Volume diffusion case. J Am Ceram Soc 79:193-198 Jung H, Karato S-i (2001a) Effects of water on dynamically recrystallized grain size of olivine. J Struct Geol 23:1337-1344 Jung H, Karato S-i (2001b) Water-induced fabric transitions in olivine. Science 293:1460-1463 Karato S-i (1986) Does partial melting decrease the creep strength of the upper mantle? Nature 319: 309-310 Karato S-i (1993) Importance of anelasticity in the interpretation of seismic tomography. Geophys Res Lett 20:1623-1626 Karato S-i (1995) Effect of water on seismic wave velocities in the upper mantle. Proc Japan Acad B 71:61-66

Seismic Wave Attenuation: Energy Dissipation in Crystalline Solids

289

Karato S-i, Karki B (2001) Origin of lateral variation of seismic wave velocities and density in the deep mantle. J Geophys Res 106:21771-21783. Karato S-i, Spetzler HA (1990) Defect microdynamics in minerals and solid-state mechanisms of seismic wave attenuation and velocity dispersion in the mantle. Rev Geophys 28:399-421 Karato S-i, Paterson MS, FitzGerald JD (1986) Rheology of synthetic olivine aggregates: Influence of grain size and water. J Geophys Res 91:8151-8176 Kohlstedt DL (1990) Chemical analysis of grain boundaries in an olivine-basalt aggregate using high resolution, analytical electron microscopy. In The Brittle Ductile Transition in Rocks: The Heard Volume. Duba AG, Durham WB, Handin JW, Wang HF (eds) American Geophysical Union, Washington, p 211-218 Kohlstedt DL, Weathers MS (1980) Deformation-induced microstructures, paleopiezometers, and differential stresses in deeply eroded fault zones. J Geophys Res 85:6269-6285 Kohlstedt DL, Keppler H, Rubie DC (1996) Solubility of water in the α, β and γ phases of (Mg,Fe)2SiO4. Contrib Mineral Petrol 123:345-357 Kê TS (1947) Experimental evidence of the viscous behavior of grain boundaries in metals. Phys Rev 71:533-546 Kê TS (1999) Fifty-year study of grain boundary relaxation. Metall Mater Trans A 30A:2267-2295 Lakes RS (1999) Viscoelastic Solids. CRC Press, Boca Raton, Florida Lakes RS, Quackenbush J (1996) Viscoelastic behaviour in indium-tin alloys over a wide range of frequencies and times. Philos Mag Lett 74:227-232 Lerner I, Chiang S-W, Kohlstedt DL (1979) Load relaxation studies for four alkali halides. Acta Metall 27:1187-1196 Lifshits IM, Shikin VB (1965) The theory of diffusional viscous flow of polycrystalline solids. Soviet Physics-Solid State 6:2211-2218 Mackwell SJ, Kohlstedt DL (1990) Diffusion of hydrogen in olivine: Implications for water in the mantle. J Geophys Res 95:5079-5088 Mackwell SJ, Kohlstedt DL, Paterson MS (1985) The role of water in the deformation of olivine single crystals. J Geophys Res 90:1319-1333 Maier J (1994) Defect chemistry at interfaces. Solid State Ionics 70/71:43-51 Mavko GM (1980) Velocity and attenuation in partially molten rocks. J Geophys Res 85:5173-5189 Mavko GM, Nur A (1975) Melt squirt in the asthenosphere. J Geophys Res 80:1444-1448 McKenzie D (1984) The generation and compaction of partially molten rock. J Petrol 25:713-765 Mei S, Kohlstedt DL (2000) Influence of water on plastic deformation of olivine aggregates, 1. Diffusion creep regime. J Geophys Res 105:21457-21469 Molodenskiy SM, Zharkov VN (1982) Chandler wobble and frequency dependence of Qμ of the Earth's mantle. Phys Solid Earth 18:245-254 Mosher DR, Raj R, Kossowsky (1976) Measurement of viscosity of the grain boundary phase in hotpressed silicon nitride. J Mater Sci 11:49-53 Nabarro FRN (1948) Deformation of crystals by the motion of single ions. In Report of a Conference on the Strength of Solids. Physical Society of London, UK, p 75-90 Nowick AS, Berry BS (1972) Anelastic Relaxation in Crystalline Solids. Academic Press, San Diego O’Connell RJ, Budiansky B (1978) Measures of dissipation in viscoelastic media. Geophys Res Lett 5:5-8. Paterson MS (1976) Some current aspects of experimental rock mechanics. Philos Trans R Soc Lond A 283:163-172 Pezzotti G, Ota K, Kleebe H-J (1996) Grain boundary relaxation in high-purity silicon nitride. J Am Ceram Soc 79:2237-2246 Plookphol T (2001) Similarity and scaling properties in dislocation microstructures generated during hightemperature load relaxation and creep of rock salt and San Carlos olivine single crystals. PhD dissertation, University of Wisconsin–Madison, Madison, Wisconsin Raj R (1975) Transient behavior of diffusion-induced creep and creep rupture. Metall Trans A 6A: 1499-1590 Raj R, Ashby MF (1971) On grain boundary sliding and diffusional creep. Metall Trans 2:1113-1127 Richter FM, McKenzie D (1984) Dynamical models for melt segregation from a deformable matrix. J Geol 92:729-740 Ricoult DL, Kohlstedt DL (1983) Structural width of low-angle grain boundaries in olivine. Phys Chem Minerals 9:133-138 Riley GN Jr, Kohlstedt DL (1991) Kinetics of melt migration in upper mantle-type rocks. Earth Planet Sci Lett 105:500-521 Romanowicz B, Durek JJ (2000) Seismological constraints on attenuation in the Earth: A review. In Earth’s Deep Interior: Mineral Physics and Tomography from the Atomic to the Global Scale. Karato S-i,

290

Cooper

Forte AM, Liebermann RC, Masters G, Stixrude L (eds) American Geophysical Union, Washington, p 161-179 Schmalzried H (1995) Chemical Kinetics of Solids. VCH, Weinheim, FRG Schoeck G, Bisogni E, Shyne J (1964) The activation energy of high temperature internal friction. Acta Metall 12:1466-1468 Scott DR, Stevenson DJ (1986) Magma ascent by porous flow. J Geophys Res 91:9283-9296 Sickafus K, Sass SL (1985) Observation of a grain boundary phase transformation induced by solute segregation. J Vac Sci Tech 3:1525-1530 Simmons G, Wang HF (1971) Single Crystal Elastic Constants and Calculated Aggregate Properties: A Handbook (Second Edn). M.I.T. Press, Cambridge, Massachusetts Smith CS (1952) Grain shapes and other metallurgical applications of topology. In Metal Interfaces. American Society for Metals, Cleveland, Ohio, p 65-108 Spiegelman M, Kenyon PM (1992) The requirements for chemical disequilibrium during magma migration. Earth Planet Sci Lett 109:611-620 Stevenson DJ (1986) On the role of surface tension in the migration of melts and fluids. Geophys Res Lett 13:1149-1152 Stone DS (1991) Scaling laws in dislocation creep. Acta Metall Mater 39:599-608 Sutton AP, Balluffi RW (1995) Interfaces in Crystalline Materials. Clarendon Press, Oxford, UK Tan B, Jackson I, FitzGerald J (1997) Shear wave dispersion and attenuation in fine-grained synthetic olivine aggregates: Preliminary results. Geophys Res Lett 24:1055-1058 Tan BH, Jackson I, FitzGerald JD (2001) High-temperature viscoelasticity of fine-grained polycrystalline olivine. Phys Chem Minerals 28:641-664 Thigpen L, Hedstrom GW, Bonner BP (1983) Inversion of creep response for retardation spectra and dynamic viscoelastic functions. J Appl Mech 105:361-366 Toomey DR, Wilcock WSD, Solomon SC, Hammond WC, Orcutt JA (1998) Mantle seismic structure beneath the MELT region of the East Pacific Rise from P and S wave tomography. Science 280: 1224-1227 Twiss RJ (1977) Theory and applicability of a recrystallized grain size paleopiezometer. Pure Appl Geophys 115:227-244 Vaughan PJ, Kohlstedt DL, Waff HS (1982) Distribution of the glass phase in hot-pressed, olivine-basalt aggregates: An electron microscopy study. Contrib Mineral Petrol 81:253-261 Waff HS (1980) Effects of the gravitational field on liquid distribution in partial melts within the upper mantle. J Geophys Res 85:1815-1825 Waff HS, Bulau JR (1979) Equilibrium fluid distribution in an ultramafic partial melt under hydrostatic stress conditions. J Geophys Res 84:6109-6114 Waff HS, Faul UH (1992) Effects of crystalline anisotropy on fluid distribution in ultra mafic partial melts. J Geophys Res 97:9003-9014 Walsh JB (1969) A new analysis of attenuation in partially melted rock. J Geophys Res 74:4333-4337 Webb S, Jackson I, FitzGerald J (1999) Viscoelasticity of the titanate perovskites CaTiO3 and SrTiO3 at high temperature. Phys Earth Planet Inter 115:259-291 Webb SC, Forsyth DW (1998) Structure of the upper mantle under the EPR from waveform inversion of regional events. Science 280:1227-1229 Xu Y, Zimmerman ME, Kohlstedt DL (2002) Deformation behavior of partially molten mantle rocks. In MARGINS Theoretical and Experimental Earth Science. Karner G (ed) Columbia University Press, New York (in press) Yan SC, Kê TS (1987) Internal friction peaks associated with the polygonization boundaries in aluminum and dilute aluminum-copper alloys. Phys Stat Solidi 104:715-721 Zener C (1941) Theory of elasticity of polycrystals with viscous grain boundaries. Phys Rev 60:906-908 Zener C (1948) Elasticity and Anelasticity of Metals. University of Chicago Press, Chicago

10

Texture and Anisotropy Hans-Rudolph Wenk Department of Earth and Planetary Science University of California Berkeley, California 94720

INTRODUCTION The study of preferred orientation of minerals in rocks dates back to Omalius d’Halloy (1833) who attributes a special significance to the alignment of crystals as an indicator of the formation process. Much later the influence of crystal alignment on physical properties was quantified (e.g., Weissenberg 1922, Voigt 1928, Reuss 1929). Only recently has this field emerged as a coherent part of earth science research linking such branches as mineralogy, petrology, structural geology, geodynamics and seismology. The reason for this was the emergence of quantitative methods to analyze preferred orientation, or “texture” as it was first called by Naumann (1850). These methods were largely developed in collaboration with materials science and mechanics. Quantitative measurements, detailed field studies, rigorous data analysis, theories to predict textures, and improvements in characterizing seismic anisotropy in the Earth are leading to a coherent picture that is now being refined. Though seismologists have long accepted that there is a causal relationship between anisotropic propagation of seismic waves, the deformation field and crystal orientation, the prevailing view is still largely the mythological concept that seismic fast directions align with the flow direction. While this may be approximately the case for olivine deformed under certain conditions, it is certainly no universal law, as we will try to illustrate in this review. The “fast” direction of a crystal depends on the mineral species and its crystal structure. The alignment of crystals depends on microscopic, intracrystalline deformation systems and the deformation history. Both relationships are complex and not intuitive, but there are well-established theories to compute single crystal physical properties as well as orientation patterns. Simulations can be compared with experimental data and then applied with some caution to the macroscopic Earth. This review is intended to provide a brief introduction, highlighting some of the issues with examples, and refer new researchers in the field of texture and anisotropy to important publications. Since the classic book of Sander (1950), there have been newer books on texture analysis in earth science (e.g., Wenk 1985; Kocks et al. 2000), as well as published collections of research papers in conference proceedings (e.g., Bunge et al. 1994), special journal issues (Leiss et al. 2000) and the tri-annual proceedings of the International Conferences of Textures of Materials (ICOTOM). These publications should be consulted for details. MEASUREMENTS OF TEXTURES Overview Interpretation of textures has to rely on a quantitative description of orientation characteristics. Two types of preferred orientations are distinguished: The lattice preferred orientation (LPO) or ‘texture’ (also ‘preferred crystallographic orientation’) and the shape preferred orientation (or ‘preferred morphological orientation’). Both can be correlated, such as in sheet silicates with a flaky morphology in schists, or fibers in fiber reinforced ceramics. In many cases they are not. In a rolled cubic metal the grain shape depends on the deformation rather than on the crystallography. Many methods have been used to determine preferred orientation. Geologists have 1529-6466/00/0051-0010$05.00

292

Wenk

extensively applied the petrographic microscope equipped with a universal stage to measure the orientation of morphological and optical directions in individual grains (e.g., Phillips 1991, Wahlstrom 1979, Wenk 1985). Metallurgists have used a reflected light microscope to determine the orientation of cleavages and etch pits (e.g., Nauer-Gerhardt and Bunge 1986). With advances in image analysis, shape preferred orientation can be determined quantitatively and automatically with stereological techniques. Optical methods of LPO measurements of some minerals have also been automated (Heilbronner and Pauli 1993, Stöckhert and Duyster 1999) Today diffraction techniques are most widely used to measure lattice preferred orientation (see e.g., Bunge 1986, Snyder et al. 1999, Wenk 2000). X-ray diffraction with a pole figure goniometer is a routine method. For some applications synchrotron X-rays provide unique opportunities. Neutron diffraction offers some distinct advantages, particularly for large bulk samples. Electron diffraction using the transmission (TEM) or scanning electron microscope (SEM) is gaining interest, because it permits one to correlate microstructures, neighbor relations and texture. There are two distinct ways to measure orientations. One way is to average over a large volume of a polycrystalline aggregate. A pole figure collects signals from many crystals and spatial information is lost (for example, misorientations with neighbors), but also some orientation relations (such as how x, y, and z-axes of individual crystals correlate). The second method is to measure orientations of individual crystals. In that case orientations and the orientation distribution can be determined unambiguously. Also, if a map of the microstructure is available, the location of a grain can be determined and relationships with neighbors can be evaluated. X-ray pole figure goniometer X-ray diffraction was first employed by Wever (1924) to investigate preferred orientation in metals, but only with the introduction of the pole-figure goniometer and use of electronic detectors did it become a quantitative method (Schulz 1949). The principle is based on Bragg’s law: in order to determine the orientation of a given lattice plane, hkl, of a single crystallite, the detector is first set to the proper Bragg angle, 2θ of the diffraction peak of interest, then the sample is rotated in the monochromatic beam until the lattice plane hkl is in the reflection condition (i.e., the normal to the lattice plane or diffraction vector is the bisectrix between incident and diffracted beam). The goniometer rotations are related to the angular coordinates that define a crystal orientation in the sample reference system. In the case of a polycrystalline sample, the intensity recorded at a certain sample orientation is proportional to the volume fraction of crystallites with their lattice planes in reflection geometry. Determination of texture can be done on a sample of large thickness and a plane surface on which X-rays are reflected (Fig. 1a, reflection) or on a thin slab which is penetrated by X-rays (Fig. 1b, transmission). Because X-rays are strongly absorbed by matter, the transmission method is generally only applicable to very thin foils or wires (= ∫ Γ c (r )A(r )d Ω.

(18)

The extent of the domain Ω depends on the specific parameterization of orientation space and the symmetries exhibited by the crystal. The macroscopic behavior is the average of the behaviors of the individual crystals. We obtain the average by an integration of the crystal quantity, weighted by the probability density, over orientation space. For an arbitrary crystal quantity, Γc, the average value, , is given by Ω

In particular, we require that the deviatoric stress and deformation rate at the macroscopic level reflect the average of their counterparts at the crystal level so that D′ = < D′c > = ∫ D′c A(r )d Ω

(19)

σ ′ = < σ ′c > = ∫ σ ′c (r ) A(r )d Ω.

(20)

Ω

and Ω

In instances in which the texture representation consists of a discrete sample of orientations taken from the orientation distribution, the macroscopic values are simply the weighted averages of the crystal quantities

Modeling Deformation of Polycrystalline Rocks: Finite Element Methods

337

D′ = < D′c > = ∑ wc D′c

(21)

σ ′ =< σ ′c >= ∑ wcσ ′c ,

(22)

c

and

c

where wc is a set of weights assigned with one-to-one correspondence to the orientations (Kocks et al. 1998). The values of weights are chosen to emphasize those crystals corresponding to regions in the orientation domain where A is high. Linking crystal responses to continuum scale motion

Recall from Equations (14) and (15) that the relationship between the deformation rate and deviatoric stress in a single crystal is an invertible relationship of the form or

D′c = Mc ⋅σ ′c

(23)

(

σ ′c = S c ⋅ D′c = M c

)

−1

⋅ D ′c ,

(24)

where M c and Sc are the single crystal viscoplastic compliance and stiffness, respectively. If we proceed with averaging these relationships we obtain and

< D′ c> = < Mc · σ ′c >

(25)

< σ ′c > = < Sc · D′c >. (26) At the macroscopic scale, we assume the deviatoric stress and deformation rate are related in an analogous form σ ′ = S · D′. (27) To relate Equations (25) and (26) with Equation (27), we must designate which hypothesis to invoke to link the microscopic (single crystal) and macroscopic behaviors. Two possibilities discussed earlier are: (1) the velocity gradient is identical from crystal to crystal over an aggregate, which can be thought of as an extension of the hypothesis of Taylor (1938), and (2) the stress is identical in each crystal (Prantil et al. 1995), which can be thought of as an extension to the hypothesis of Sachs (1928). For the first case we note in Equation (26), the crystal deformation rate can be removed from the orientation average if it is identical for all orientations to give σ ′ = < σ ′c > = < Sc · D′c > = < Sc > · D′ = S · D′. (28) Using this assumption renders equal straining in all crystals of an aggregate. For the second case, the identical stress may be taken outside the orientational average to give D′ = < D′c > = < Mc ⋅ σ ′c > = < Mc > ⋅ σ ′c = < Mc > ⋅ σ ′. (29) c To obtain the form of Equation (27), < M > is inverted σ ′ = < Mc >−1 ⋅ D′ = S · D′. (30) Estimating S by these two approaches gives upper and lower bounds of the stiffness, respectively, for a specific texture. With the Taylor hypothesis, compatibility is enforced, but equilibrium may be violated. With the equal stress constraint, equilibrium is satisfied identically, but the crystal deformations may not be compatible. SMALL-SCALE HYBRID ELEMENT FORMULATION

Small-scale formulations are performed using a hybrid finite element formulation. Central to hybrid formulations is the concept of domain decomposition, which here

338

Dawson

coincides with crystals of a polycrystal. The stress is represented in a manner that satisfies equilibrium within the domain a priori so that enforcement of equilibrium for the full polycrystal consists of matching tractions across the domain (crystal) interfaces. To this end, the traction equilibrium residual, after utilizing the divergence theorem, is given as

(

)

Rt = ∑ ⎡ ∫ σ ′c − pI : ∇ ( Φ ) : d Ω − ∫ Φ ⋅ t d Γ ⎤ = 0 , ⎢ Ω ⎥⎦ Γe e ⎣ e

(31)

where t is the surface traction, p is the pressure, I is the identity tensor, Φ are vector weights and Γe is the surface bounding Ωe. Using Equation (13), which gives a linearized version of the relation that quantifies intracrystal slip, a residual is constructed over the each subdomain volume, Ωe

(

)

Rσ = ∫ Ψ ⋅ M c : σ' c − D c d Ω = 0, Ωe

(32)

where Mc serves as a crystal viscoplastic matrix that relates the crystal deviatoric stress, σ ′c, to the crystal deformation rate, Dc, and Ψ are weighting functions. Consistent with the order of Equation (13), Ψ are second order quantities. Finally, allowing for slight compressibility a penalized residual on the incompressibility condition is constructed for each domain using scalar weights, ϒ 1 ϒdiv u d Ω = 0, (33) Ωe ε ∫Ωe where ε is a penalty parameter. Trial functions are introduced in the residuals for the interpolated field variables. In the hybrid formulation, both the velocity and the stress (deviatoric and spherical portions) are represented with trial functions Rv = ∫ ϒp d Ω +

σ ′c = NsSh; p = NpPh; u = NuUh .

(34) Here, N , N , and N are interpolations functions and S , P , and U are nodal quantities for each variable being represented. For the deviatoric stress, piecewise discontinuous functions are specified which satisfy equilibrium within an element a priori (Bratainu and Atluri 1983). The weights are constructed according to a Galerkin methodology as s

p

u

h

h

h

Ψ = Ns Ψh; ϒ = Np ϒh; Φ = Nu Ψh . (35) Each residual is written in matrix form using the trial and weight functions. For the crystal slip relation, Equation (32), nodal stresses are related to the nodal velocities H ⋅ Sh – RT ⋅ Uh = 0 ,

(36)

H = ∫ N sT ⋅ M c ⋅ N s d Ω

(37)

R = ∫ BT ⋅ N s d Ω ,

(38)

where Ωe

and Ωe

where B contains spatial derivatives of Nu such that (39) D = BUh and D is a vector containing the components of the deformation rate. Solving for the nodal stress values gives Sh = H−1 ⋅ RT ⋅ Uh = 0. In a similar way, the incompressibility constraint is written in matrix form as

(40)

Modeling Deformation of Polycrystalline Rocks: Finite Element Methods

339

1 M p ⋅ P h = − GT U h , ε

(41)

M p = ∫ N pT N p d Ω

(42)

G = ∫ BT hT N p d Ω.

(43)

where Ω

and Ω

Here, h is the trace operator. It is possible to solve for Ph over each domain independently, giving 1 P h = − M p −1 ⋅ G T U h . (44) ε The traction residual becomes

∑ ( R ⋅ S h − G ⋅ P h − f ) = 0.

(45)

e

Substituting Equations (40) and (44) into Equation (45) for Sh and Ph, repectively, give

∑ ( ⎡⎣ R ⋅ H −1 ⋅ RT − G ⋅ M p−1 ⋅ GT ⎤⎦ ⋅U h − f ) = 0,

(46)

f = ∫ N uT ⋅ t d Γ.

(47)

e

where Γe

This discretized traction residual is solved simultaneously for the velocity field, from which the stress may be recovered. LARGE-SCALE VELOCITY-PRESSURE FORMULATION

For large-scale applications, the finite element formulation is similar to traditional developments for viscous flow (Mathur and Dawson 1989). The linear momentum and conservation of mass balance laws are used to form the residuals for the finite element approximations constructed to evaluate the velocity fields. Coupling to the solution of the temperature distribution (Zienkiewicz and Taylor 1989). is straightforward and will not be discussed here. A weighted residual for equilibrium (Johnson 1987) is written over the body volume, Ω, using vector weights, Φ as

(

)

Ru = ∫ Φ ⋅ divσ T + b d Ω = 0. Ω

(48)

This is modified to obtain the weak form through integration by parts and application of the divergence theorem to give

(

)

Ru = − ∫ tr σ 'T ⋅∇Φ d Ω + ∫ pdivΦd Ω + ∫ t ⋅ Φd Γ + ∫ b ⋅ Φd Ω =0, Ω

Ω

Γ

Ω

(49)

where the Cauchy stress has been divided into its deviatoric (σ′) and spherical parts (pI). The body force is b and t is the traction acting on the boundary Γ of Ω. For an incompressible medium trD = divu = 0, (50) which can be enforced using a consistent penalty approach (Engelman et al. 1982). First a residual is written using the incompressibility constraint (Eqn. 50) Rv = ∫ ϒdivud Ω = 0, Ω

(51)

340

Dawson

where ϒ are scalar weights. Slight compressibility is admitted, giving the possibility to write a penalty constraint in a form consistent with the finite element discretization Rv = ∫ ϒpd Ω + Ω

1 ϒdivud Ω = 0, ε ∫Ω

(52)

where ε is a penalty parameter. Finite element interpolation is introduced as u = NuUh

p = NpPh

and

(53)

for the trial functions and Ψ = NuΨh and ϒ = NpΦh

(54) u

p

for the weighting functions, where N and N are functions of the position x. The trial functions for u are continuous over element boundaries, but for p discontinuous interpolation is specified to match the interpolation for the gradient of the velocity. With discontinuous pressure interpolation, the penalized incompressibility constraint can be written element-by-element as 1 Ph = − M p −1G T U h , ε

(55)

M p = ∫ N pT N p d Ω

(56)

G = ∫ BT hT N p d Ω.

(57)

where Ω

and Ω

Here B is defined in Equations (39) and h again is the trace operator. We introduce the mechanical behavior from Equation (27) in matrix form as σ ′= SD′. (58) Substitution of this relationship into Equation (49) eliminates the deviatoric stress. Taken with Equation (55) for eliminating the pressure, the residual is written for arbitrary variations in the weights as 1 ⎛ p −1 T ⎞ h ⎜ K s + GM G ⎟ U = F , ε ⎝ ⎠

(59)

K s = ∫ B T S Bd Ω

(60)

F = ∫ N T td Γ + ∫ N T bd Ω.

(61)

where Ω

and Γ

Ω

From Equation (59), the velocity field is determined for a specific combination of geometry, loading, and state. HALITE TEXTURE EVOLUTION: A SMALL-SCALE APPLICATION Generalities

Modeling the plastic deformations and associated texture evolution in halite is challenging due to its high level of plastic anisotropy, at least at low homologous temperature. Small-scale finite element modeling provides an avenue to study how the anisotropy influences the degree of heterogeneity of deformation at the crystal scale and how this in turn

Modeling Deformation of Polycrystalline Rocks: Finite Element Methods

341

Figure 1. The finite element meshes of the inner 1000 crystals of an aggregate of halite crystals corresponding to the initial and deformed shapes (repaired). Gray shades show the longitudinal component of the deformation rate.

affects the texture evolution. Although halite possesses a cubic structure, its response under mechanical load is highly anisotropic because of the sparseness of available slip systems. At the level of a single crystal, some modes of deformation require considerably higher stress levels than others. As a consequence, there is a tendency in polycrystals for crystals that are more favorably oriented to deform more than others that are less favorably oriented. A consequence of this deformation heterogeneity is that assuming that all grains deform identically when modeling the mechanical response leads to poor texture predictions. As example of the use of the small-scale formulation we show results reported in (Lebensohn et al., in press) in which predictions obtained using finite element simulations, self-consistent simulations and Taylor simulations were compared. The intent of the simulations is to learn more about how straining is partitioned among crystals in an aggregate and from there to better understand how texturing proceeds under continued deformation.

Figure 2. Measured and computed stress strain curves for halite deformed in extension at 20°C and 100°C computed using the finite element model, a Taylor assumption, and a viscoplastic model.

342

Dawson

Simulation specifics Simulations were performed on a sample of halite consisting of 4096 crystals. The starting texture corresponded to a uniform orientation distribution, with orientations randomly chosen from the distribution and assigned to the crystals comprising the sample. The initial shape of all crystals (and sample) was a cube. Boundary conditions were imposed to replicate conditions corresponding to an extension experiment with constant true strain rate. As the deformation proceeded, the crystals quickly became distorted and irregular due to the strain heterogeneity. To compensate for this, the crystals periodically were given a brick shape corresponding to the current exterior shape of the sample without altering the lattice orientations as shown in Figure 1. The slip system parameters were chosen so that the computed average stress, as determined by the integrated surface tractions, matched measured stress strain curves, as shown in Figure 2.

Figure 3. Tensile component of the deformation rate at high temperature (100°C) for the halite crystals under three modeling assumptions.

Simulation results

A goal of modeling is to accurately capture the heterogeneity of deformation among the crystals of an aggregate. A tradeoff exists between the complexity of the model and its ability to capture this heterogeneity. Modelers seek a balance such that essential features of a deformation are accurately computed with unnecessary details and accompanying computational burdens. To help understand where this balance lies, we compare the variation in deformation rate over the aggregate computed with the finite element model (HEP) to two other modeling assumptions: a Taylor assumption (FC, indicating full kinematic constraints) and a viscoplastic self-consistent formulation (VPSC). Figure 3 shows how the deformation rate varies as a function of crystal orientation as quantified with the Taylor factor. The FC model, by definition, has no variation. The VPSC model has a strong systematic dependence. The HEP model has weaker dependence with Taylor than the VPSC overall, but greater variation of deformation at fixed Taylor factor. This stems from the unique neighborhood that each finite element has, resulting in different deformation rates for two crystals of similar orientation but different surroundings. This difference in local behavior of the grains influences the evolution of properties with continuing deformation. The lattice orientations evolve with deformation according to the equations for crystal plasticity reviewed in the prior sections of the paper. Textures after various amounts of deformation are shown as inverse pole figures of the extension direction in Figure 4 for the three modeling assumptions. As is evident, certain component of the texture are elevated and other suppressed by the degree to which the deformations can vary over the polycrystal. Texture development is strongest for the Taylor (FC) model

Modeling Deformation of Polycrystalline Rocks: Finite Element Methods

343

and weakest for the finite element (HEP) model. In all cases, a strong 111 component develops, but for the VPSC a 100 component also is significant.

Figure 4. Computed inverse pole figures for texture evolution of halite at 100°C for increasing amount of extension. Directions in inverse pole figures are: lower left – 001; lower right – 011; upper right – 111.

MANTLE CONVECTION – A LARGE-SCALE APPLICATION Generalities

Of particular importance for processes in the crust, such as volcanic activity, earthquakes, and mountain building, is the deformation of the mantle. Within the mantle, large cells of convection are induced by instabilities and driven by temperature and density gradients. The convective flow of the mantle substantially alters the internal structure, and the mechanical properties derived from it. The main constituent of the upper mantle is olivine, which exhibits about a 25% difference in acoustic speed between the slowest and the fastest crystal directions. Due to this difference, a polycrystal with preferred orientations of component crystals will display a directional dependence (anisotropy) of seismic wave propagation. Indeed, seismic waves do travel faster (about 5%) perpendicular to the oceanic ridges than parallel to them (Morris et al. 1969). As early as 1964, Hess interpreted this as a result of a preferential alignment of crystals with directional properties, and proposed that this alignment was attained during mantle convection (Hess 1964). Chastel and coworkers presented model for predicting the development of anisotropy in a convection cell and discussed implications for convection in the Earth’s mantle (Chastel et al. 1993). The results indicate that during convection, crystals reorient into characteristic textures that cause anisotropic physical properties. It was observed that these patterns are highly heterogeneous over the span of the convection cell and depend on the specific straining history along individual streamlines of the cell. The rotations that reorient crystal lattices are due to activation of slip systems in deforming single crystals as well as the macroscopic rotation of material progressing along a streamline. Both causes of lattice reorientation are most intense during upwelling at ridges and during subduction. Thus, texture development is particularly strong in these regions. Using the crystallographic texture throughout the convection cell and the physical properties of single crystals, seismic

344

Dawson

Figure 5. Schematic diagram of the convection cell showing the upper and lower regions of the mantle and the crust.

velocity averages over mantle sectors were computed and the directional dependence of wave propagation determined. The predicted azimuthal variation of p-wave anisotropy of 510% reported in that publication agrees well with actual observed values. The convection cell used by Chastel et al. was highly idealized. In a later investigation reported by Dawson and Wenk (2000), a more realistic mantle was modeled, one with dimensions corresponding to those in the Earth and with material properties that are closer to those which are believed to exist. The convective cell was discretized with a greater number of finite elements to better resolve gradients of the flow and use a larger number of crystals to represent each polycrystal. Apart from that, the procedure was similar to Chastel’s. The ‘mantle’ was still idealized: the model was two-dimensional; there was no interaction with the crust (e.g., during subduction); and the sole phase change (at 650 km) affected only the mechanical properties. Above 650 km, the mantle was assumed to be solely composed of Table 1. Parameters used in the simulations (Turcotte and Schubert 1990; Blankenbach 1989). If ranges are given, the first value is for the top of the mantle and the second for the bottom. Quantity Density Heat capacity Thermal conductivity Volumetric thermal expansion Gravity Youngs modulus Compressibility Kinematic viscosity Dynamic viscosity (μ = νρ)

Symbol ρ Cp k α g E β ν μ

Units kg/m-3 J/(kgK) W/(mK) 1/K m/s-2 Pa 1/Pa m2/s Pa s

Value 4(3.4-5.6)×103 1.25×103 5 2.5×10-5 9.8(9.86-10.68) 1.6-7.6×1011 7.5(8.0-1.5)×10-12 2.5×1017-19 1×1021-23

Modeling Deformation of Polycrystalline Rocks: Finite Element Methods

345

Figure 6. Left: temperature distribution during mantle convection showing the effects of upwelling (on the left) and subduction (on the right). Right: convective velocity field shown using scaled vectors directed tangent to the flow.

olivine. Below 650 km, the mantle was assumed to be isotropic with a viscosity higher that that of the upper mantle. The model was intentionally kept simple to emphasize the effects of the convective flow on the internal structure of the mantle material and on the resulting anisotropy in its mechanical properties. Here we summarize some of those results. Simulation specifics The convection cell is 2900 km deep and 3000 km wide. Within this region are three zones: the lower mantle extends from 650 km to 2900 km depth across the width of the cell; the upper mantle extends from the bottom of the crust to the lower mantle, also across the full width of the cell; the crust extends from the surface to the top of the upper mantle, varying in thickness from 50 km (along the boundary with upwelling) to 125 km (along the boundary with subduction).

The cell geometry is shown schematically in Figure 5. Along the bottom of cell (lower extent of the lower mantle), no vertical velocity is allowed. Along both lateral surfaces, no horizontal velocity is allowed. The crust is assumed rigid, and at the boundary between the upper mantle and the crust the velocity of the upper mantle is required to be tangential to the crust/upper mantle boundary. It is not necessary to model the crust explicitly with these boundary conditions, and so only the upper and lower mantle zones have been discretized with finite elements. The top surface of the mantle is required to be 1600 K; the bottom surface of the lower mantle is fixed at 2900 K. The lateral boundaries are adiabatic. A total of 240 higher order, isoparametric elements are used to discretize the full mantle, 120 in the upper mantle and 120 in the lower mantle. An Eulerian procedure is employed, so the elements are spatially fixed and mantle material flows though them. The driving force for the flow arises from the density gradients associated with the nonuniform temperature field. To accommodate this, an iterative procedure is used to couple the heat transfer solution with the solution for the velocity field. Table 2. Single crystal elastic moduli (GPa) (Simmons and Wang 1971). Voigt notation. C11

C12

C13

C22

C23

C33

C44

C55

C66

324

59

79

249

78

249

66.7

81

79.3

346

Dawson

Figure 7. 100 pole figures along one streamline illustrating the evolution of texture for a rock traveling near the cell perimeter. 100 pole figures A-E correspond to positions A-E, respectively.

The macroscopic material stiffness given by Equation (27) is derived as the average of single crystal responses using either an upper or lower bound assumption. The slip system response for the single crystals is determined by matching laboratory data for compression of polycrystalline olivine, as discussed earlier. In the previous study (Chastel et al. 1993), the lower bound stiffness was used in the finite element equations and convective velocity fields were computed for two different sets of boundary conditions. Each set restricted the magnitude of the velocity to be consistent with observed convection rates. In the later study by Dawson and Wenk (2000), the anisotropic stiffness derived this way was replaced with an isotropic stiffness based on a single viscosity, to avoid having to place an explicit constraint on the magnitude of the velocity field. These isotropic properties were chosen in accordance with the geodynamics literature (Turcotte and Schubert 1990) and benchmark calculations (Blankenbach 1989). Corresponding values are listed in Table 1. Values used in the present simulations correspond to the center of this range. The reason for making this change was that the viscosity implied by the anisotropic stiffness as derived from experiment is orders of magnitude lower than that inferred from convective rates or glacial rebound. While anisotropy was neglected for the purposes of computing the velocity field, the simulations still based the texturing on the polycrystal model, including the single crystal behavior for olivine and the lower bound linking hypothesis. In this respect coupling between the macroscopic velocity field and the evolving microstructure was only from the velocity field to the microstructure, and not from the microstructure to the velocity field. Simulation results

Figure 6 illustrates the temperature distribution and the velocity field when convection has reached a steady state, displaying patterns dominated by convective heat exchange. The evolution of texture is best evaluated by tracing material flow along various streamlines as shown in Figure 7. In the lower mantle the material is assumed to be isotropic. When it enters the upper mantle and transforms to olivine, the texture evolution begins starting from an assumed uniform (untextured) orientation distribution as was assumed in (Chastel et al. 1993; Dawson and Wenk 2000). Strong preferred orientation (represented

Modeling Deformation of Polycrystalline Rocks: Finite Element Methods

Figure 8. Left: 100 pole figures displayed at the centroids of the finite elements of the upper mantle (not to scale). Right: longitudinal p-wave propagation surfaces at the centroids of the finite elements of the upper mantle corresponding to the texture patterns. Linear contours with contour interval 100 m/s. Dot pattern below 7200 m/s. [Note: The plot was rotated by 90° counterclockwise on the page to facilitate display.] Horizontal Earth direction is vertical on the page with the surface of the Earth to the left. Upwelling occurs at the left side of the rotated diagram, spreading is toward the right, and downwelling is on the right side.

347

348

Dawson

as 100 pole figures) develops during upwelling with high shear. The pattern changes as the flow lines change direction from vertical to horizontal. During horizontal spreading orientation changes are minimal. Finally, during subduction, a new pattern develops which is generally weaker but does not revert to randomness at the interface with the lower mantle. In our model the time for the convection path of an outer streamline in the upper mantle is about 120 million years, corresponding to a spreading velocity of 2 cm/yr. The texture patterns for the element centroids of the upper mantle are presented in Figure 8 as 100 pole figures. Great heterogeneity is observed, vertically and laterally, with regions having strong texture while others have almost random orientation distributions. The uppermost layer is strongly textured, but immediately underneath there is a fairly isotropic layer. The center of the cell is again strongly textured. All pole figures display a statistical diadic symmetry, consistent with the two-dimensional (plane strain) deformation. The pole figures have no mirror planes because the deformation has a simple shear component (texture calculations and representations were done with BEARTEX (Wenk et al. 1998).

Figure 9. Seismic p-wave velocity variations in regions of upwelling and spreading.

If the orientation distribution and the tensor properties of the single crystal are known, then the physical properties of the textured polycrystal can be calculated by appropriate averaging. In the case of the mantle, elastic properties are of main interest. The single crystal elastic moduli at ambient conditions are given in Table 2 for olivine. Using these values, averaging over all 1000 crystallites in each cell was performed with the self consistent method assuming spherical grain shape (Kocks et al. 1998). From the averaged elastic constants and the density (4.5 g/cm for the central region of the mantle) propagation surfaces for longitudinal (p) and transverse (s) waves were calculated, with those for p-waves shown for all element centroids in Figure 8. The velocity maps are shown in equal area projection. We are aware that both elastic constants and density change with temperature and pressure, but in our representation we used constant values to emphasize the effects of anisotropy. All deviations from the value 7280 m/s for an isotropic polycrystal are due to texture. Similar to pole figures in Figure 8, there are large deviations from isotropy and also a heterogeneous distribution of anisotropy over the convection cell. Since the flow pattern is two-dimensional, the deformation corresponds to

Modeling Deformation of Polycrystalline Rocks: Finite Element Methods

349

plane strain, and textures and velocity surfaces have statistically a symmetry plane and a two-fold rotation axis perpendicular to it. Overall, there is a correspondence between fast velocities and 100 concentrations, since 100 is the fast single crystal direction. If earthquake travel times were recorded from waves which pass through the upper mantle at high angles, zones with high anisotropy would mainly appear as low velocity zones, since fast 100 directions are dominantly horizontal.

Figure 10. Seismic p-wave velocity distributions over the upper mantle corresponding to horizontal, vertical and normal (transverse) propagation directions. Note that in the original report of Dawson and Wenk (2000) the gray shades were in error (inverted) for two directions.

In addition to the local averages, averages have been computed over larger scales, particularly for regions of upwelling and spreading (see Fig. 9). This figure shows the azimuthal variation in p-wave velocity, as would be observed with surface waves (horizontal). The velocity variation is 5-10%, in good correspondence with observed data from Hawaii (Morris et al. 1969), the Mendocino ridge (Raitt et al. 1969) and the Southern Pacific (Nishimura and Djewonski 1984). Finally, contour maps of vertical, horizontal and normal velocities (again, assuming constant density and single crystal properties) are presented for the full convective cell. This represents the local velocities if corresponding seismic waves were propagating in the corresponding directions (see Fig. 10a-c). Due to the two-dimensional approach deviations from average in the normal direction are minimal. However, in the other two maps there are again variations of 5-10% with patterns resembling seismic tomography (Anderson and Djewonski 1984).

350

Dawson SUMMARY

Anisotropy at the level of individual crystals strongly influences the mechanical response of minerals. Polycrystal plasticity provides the theoretical basis for modeling this anisotropy. Using polycrystal plasticity it is possible to compute the dependence of the mechanical properties on the crystallographic texture and to evolve the texture with deformation. The finite element method provides a powerful tool that can be used in conjunction with polycrystal plasticity to solve complex boundary value problems. These can be at a macroscopic scale, as in the case of mantle convection, or crystal scale, as in the detailed simulation of crystal aggregates. The potential for a more fundamental understanding of mineral deformation is open through coordinated efforts of experiments and simulations using these tools. ACKNOWLEDGMENTS

This work has been supported in part by the Cornell Theory Center. The author thanks Chris Pelkie of Cornell Theory Center for assistance in the graphic presentation of the simulation results, Donald Boyce of Cornell University for help with the texture evolution computations, and Carlos Tomé of Los Alamos National Laboratory for assistance with the seismic velocity computations using the self-consistent approach. REFERENCES Anderson DL, Djewonski AM (1984) Seismic tomography. Sci Am 251:60-68 Asaro R, Needleman A(1985) Texture development and strain hardening in rate-dependent polycrystals. Acta Metall 23:923-953 Beaudoin AJ, Dawson PR, Mathur KK, Kocks UF (1995) A hybrid finite element formulation for polycrystal plasticity with consideration of macrostructural and microstructural linking. Intl J Plasticity 11:501-521 Blankenbach B (1989) A benchmark comparison for mantle convection codes. Geophys J Intl 98:23-38 Bratianu C, Atlui S (1983) A hybrid finite element method for stokes flow: Part I—formulation and numerical studies. Computer Methods Appl Mech Eng 36:23-30 Chastel YB, Dawson PR, Wenk H-R, Bennett K (1993) Anisotropic convection with implications for the upper mantle. J Geophys Res 98:17,757-17,771 Dawson PR, Marin EB (1998) Computational mechanics for metal deformation processes using polycrystal plasticity. In van der Giessen E et al. (eds) Advances in Applied Mechanics 34:78-169 Dawson PR, Wenk H-R (2000) Texturing the upper mantle during convection. Philos Mag A 80:573-598 Engelman M, Sani R, Gresho P, Bercovier M (1982) Consistent vs. reduced integration penalty methods for incompressible media using several old and new elements. Intl J Numer Methods Fluids 2:25 Hess HH (1964) Seismic anisotropy of the uppermost mantle under oceans. Nature 203:629-640 Johnson C (1987) Numerical Solution of Partial Differential Equations by the Finite Element Method. Cambridge University Press, Cambridge, UK Kocks UF, Tome CN, Wenk H-R (1998) Texture and Anisotropy. Cambridge University Press, Cambridge, UK Lebensohn R, Dawson PR, Kern H, Wenk H-R (in press) Heterogeneous deformations and texture development in halite polycrystals: comparison of different modeling approaches and experimental data. Tectonophysics Mathur KK, Dawson PR (1989) On modeling the development of crystallographic texture in bulk forming processes. Intl J Plasticity 5:67-94 Morris GB, Raitt RW, Shor GG (1969) Velocity anisotropy and delay time maps of the mantle near Hawaii. J Geophys Res 74:4300-4316 Nishimura C, Forsyth D (1989) The anisotropic structure of the upper mantle in the pacific. Geophys J 96:203-229 Parks DM, Ahzi S (1990) Polycrystalline plastic deformation and texture evolution for crystals lacking five independent slip systems. J Mech Phys Solids 38:701-724 Prantil V, Dawson PR, Chastel YB (1995) Comparison of equilibrium-based plasticity models and a Taylorlike hybrid formulation fo deformation of constrained crystal systems. Modeling Simul Mater Sci Eng 3:215-234 Raitt RW, Shor GG, Francis TJG, Morris GB (1969) J Geophys Res 74:3095-3109 Sachs G (1928) Zur Ableitung einerFliessbedingung. Z Verein Deut Ing 72:734-736

Modeling Deformation of Polycrystalline Rocks: Finite Element Methods

351

Simmons G, Wang H (1971) Single Crystal Elastic Constants and Calculated Aggregate Properties. M I T Press, Cambridge, Massachusetts Taylor G (1938) Plastic strains in metals. J Inst Metals 62:307-324 Turcotte DL, Schubert G (1990) Geodynamics Applications of Continuum Physics to Geological Problems. Wiley, New York Wenk H-R (1985) Preferred Oreintations of Deformed Metals and Rocks: An Introduction to Modern Texture Analysis. Academic Press Wenk H-R, Mathies S, Donovan J, Chateigner D (1998) Beartex: A Windows-based program system for quantitative texture analysis. J Appl Crystallogr 31:262-272 Zienkiewicz O, Taylor R (1989) The Finite Element Method. McGraw-Hill, London

12

Seismic Anisotropy and Global Geodynamics Jean-Paul Montagner(1,2,3) and Laurent Guillot(1) (1)

Seismological Laboratory, CNRS URA 195 Institut de Physique du Globe, Paris, France (2) On leave at: Seismological Laboratory California Institute of Technology, 252-21 Pasadena, California 91125 (3)

On leave at: Jet Propulsion Laboratory California Institute of Technology 4800 Oak Grove Drive, Pasadena, California 91109

INTRODUCTION For many years, seismic anisotropy was often neglected, mostly because of the inherent heavy mathematical and computational tools needed to describe and model its effects on seismic waves. The usual basic knowledge about propagation in isotropic media cannot easily apply to anisotropic media, where new phenomena come up, such as birefringence (or shear-wave splitting), or difference between directions of propagation of phase velocity and of group velocity. Consequently, geophysicists often claimed that it was a second-order effect, and considered the Earth as isotropic. This hypothesis was assumed to be a good approximation, because of the random orientation of crystals in most parts of the Earth, and of the random sampling of anisotropic regions by seismic rays. This assumption furthermore made easier the description of wave propagation, as well as the parameterization of media in inverse problems. An isotropic elastic medium can be described by two independent elastic parameters (λ and μ Lamé parameters), but the simplest anisotropic medium (transverse isotropy with a vertical symmetry axis) requires 5 independent parameters (Love 1927, Anderson 1961). To date, seismic observations have been explained in terms of isotropic (and often thermal) lateral heterogeneities, ignoring manifestations of anisotropy. However, since the 1960s, it was recognized that most parts of the Earth are not only laterally heterogeneous but also anisotropic and that seismic anisotropy provides a simple explanation of different observational data: • azimuthal variation of Pn-velocities below oceans (discovered in the 1960s; Hess (1964) explained it by anisotropy), • Rayleigh-Love wave discrepancy: It is impossible to simultaneously explain Rayleigh and Love wave dispersion by an isotropic model (Anderson 1961, Aki and Kaminuma 1963, Mc Evilly 1964), • Shear-wave splitting (or birefringence), the most unambiguous observation of anisotropy, particularly for SKS waves (Vinnik et al. 1984). Seismic anisotropy, contrary to isotropy, is reflecting some inherent organization of the matter. Different geophysical fields are involved in the investigation of the manifestations of anisotropy of Earth materials: mineral physics and geology for the study of the microscopic scale, seismology and geodynamics for scales larger than typically one kilometer. The origin of seismic anisotropy is non-unique. In the crust, the crack distribution seems to play a major role (Crampin and Booth 1985). In the upper mantle, it is usually explained by the lattice-preferred orientation of α-olivine (Nicolas and Christensen 1987, Zhang and Karato 1995) and is related to plate-tectonic processes. More generally the intrinsic anisotropy of minerals (olivine and to a less extent orthopyroxene and clinopyroxene) associated with their lattice-preferred orientation may induce large1529-6466/00/0051-0012$05.00

354

Montagner & Guillot 

scale observable and unambiguous effects, either on body waves (S-wave splitting observed on SKS (Vinnik et al. 1984), P-wave anisotropy (Babuska et al. 1984) or surface waves through the azimuthal anisotropy (Forsyth 1975) and the radial (improperly named “polarization”) anisotropy (Schlue and Knopoff 1977). These different measurements of anisotropy performed at different spatial scales were difficult to reconcile. Investigating deeper lower mantle anisotropy is a formidable challenge, because the anisotropy signal is small, often masked by upper mantle anisotropy, and its physical explanation is controversial. Most of the lower mantle seems to be isotropic (Meade et al. 1995), except the D′′-layer (Vinnik et al. 1989b, Maupin 1994) where anisotropy could result from the layering of old subducted slabs and/or melted materials. Even deeper, anisotropy has also been found in the inner core from free oscillations (Woodhouse et al. 1986) and from the P-wave travel times reported in the ISC bulletins (Morelli et al 1986). The origin, amplitude and mechanisms creating the anisotropy in the core are still subject of controversy (e.g., see Singh et al. 2000). Since these early observations of seismic anisotropy, a large and rapidly growing number of studies have confirmed its existence in the different depth ranges of the Earth. A complete and exhaustive review of these studies is beyond the scope of this paper, and, we will only underline their geodynamic implications. From the global geodynamics point of view, seismic anisotropy has many applications, although it is still in its infancy. We will show how it makes it possible to define the root of continents and to investigate the coupling between the lithosphere and the rest of the mantle (Montagner and Tanimoto 1991, Silver 1996), and more generally to gain insight into mantle convection (Anderson and Regan 1984, Montagner 1994). Mantle convection characterized by a high Rayleigh number is highly chaotic and numerical modeling demonstrates that most of the deformation takes place in boundary layers. Conversely, since seismic anisotropy is closely related to large-scale deformation (Nicolas and Christensen 1987, Karato 1989), boundary layers can be detected by the existence of seismic anisotropy (Montagner 1998, Karato 1998). We present in this paper the different depth ranges in the mantle, where seismic anisotropy was detected, i.e., uppermost mantle, transition zone and D′′-layer. In the top boundary layer of the mantle, seismic anisotropy can be directly compared to geological observations. The robust features of these different investigations in different depth ranges are presented and the enormous scientific potential of seismic anisotropy is emphasized. CAUSES OF SEISMIC ANISOTROPY FROM MICROSCOPIC TO LARGE SCALE Many processes can give rise to seismic anisotropy, and we must be very careful when interpreting it. They can be related either to anisotropic structural settings in rocks, or to the intrinsic anisotropy of minerals in the Earth. Shape Preferred Orientation (S.P.O.) Anisotropic spatial organization of matter (even though isotropic) can induce seismic anisotropy: • a fine layering of heterogeneous rocks (even though isotropic) is “seen” as transversely isotropic by a wave whose wavelength is larger than the typical thickness of layers (Backus 1962). Some authors proposed that this kind of structure, observed in ophiolites for instance, could result from the progressive mixing of rocks with heterogeneous mechanical properties and could explain the seismic anisotropy in the Earth’s upper mantle (e.g., the “marble cake” of Allègre and Turcotte 1986); • distribution of cracks (Crampin and Booth 1985) and/or fluid inclusions can induce

Seismic Anisotropy and Global Geodynamics

355

anisotropy (probably an important effect in the crust and in the inner core). Lattice Preferred Orientation (L.P.O.) Some major mineral phases of the Earth’s mantle are anisotropic in elasticity. Under some conditions, plastic deformation of these minerals can result in a preferential orientation of their lattices. This phenomenon is often considered as the origin of the large-scale seismic anisotropy in the upper mantle. We present below in a very simplified manner the conditions required to develop such a seismic anisotropy. A complete discussion of these different mechanisms at different scales can be found in recent extensive and well-documented review papers (Kendall 2000, Mainprice et al. 2000, Savage 1999). Intrinsic anisotropy of minerals. Some minerals present in the upper mantle are strongly anisotropic (see Mainprice et al. 2000 and references therein for an exhaustive review of the minerals and their anisotropic properties). The difference of P-wave velocity between the fast axis and the slow axis is larger than 20% for olivine, the main constituent of the upper mantle. Other important constituents such as orthopyroxene or clinopyroxene are anisotropic as well (>10%) (see Anderson 1989, or Babuska and Cara 1991). Some other constituents such as garnet display a very small intrinsic anisotropy. With increasing the depth, most of minerals undergo a series of phase transformations. There is some tendency (though not systematic), that with increasing pressure (whose effect is predominant in the mantle), the crystallographic structure evolves towards a more closely packed, more isotropic structure, such as cubic structure. For example, olivine transforms into β-spinel and then γ-spinel in the upper transition zone (410 to 660 km depth) and into perovskite and magnesiowüstite in the lower mantle; but perovskite (Mg,Fe)SiO3 and the pure end-member of magnesiowüstite MgO are still anisotropic. That could explain the observed anisotropy in some parts of the lower mantle (Karato 1998, or MacNamara et al. 2002). Efficient mechanisms of orientation of crystals. In order to observe seismic anisotropy, the crystals must be sensitive to the strain field, and a lattice preferred orientation must develop either from dislocation creep (activation of slip systems) or from dynamic recrystallization (see Karato (1989) and Poirier (1985), for a phenomenological description, and Ribe (1989) and Kaminski and Ribe (2001), e.g., for numerical modeling). Through the mechanisms of lattice-preferred orientation, it is found that the anisotropy of an aggregate of many minerals can be very large (Nicolas and Christensen 1987). In the lower mantle super-plasticity may be the predominant deformation mechanism, and that may cause the absence of large scale seismic anisotropy in this part of the Earth (Karato 1998). Anisotropy of assemblages of minerals. Mantle rocks are assemblages of different minerals which are more or less anisotropic. The amount of anisotropy is largely dependent on the composition of the aggregates. The relative orientations of crystallographic axes in the different minerals must not counteract in destroying the intrinsic anisotropy of each mineral. For example, the anisotropy of peridotites, mainly composed of olivine and orthopyroxene, is affected by the relative orientation of their crystallographic axes (Christensen and Lundquist 1982). According to their observations, the fast axis of olivine is parallel to the intermediate axis of orthopyroxene in the shear plane and parallel to the flow direction, but the fast axis of orthopyroxene and the slow axis of olivine are orthogonal to the shear plane (Fig. 1). The resulting anisotropy for Pwaves, however, is larger than 10% (Peselnick and Nicolas 1978). Such a large anisotropy is consistent on intermediate scale, for instance in massifs of ophiolites, which present such an anisotropy over several tens of kilometers (Nicolas 1993, Vauchez and Nicolas 1991).

356

Montagner & Guillot 

Figure 1. Figure of Christensen and Lundquist (see Anderson 1989) for olivine and orthopyroxene orientation). This observation is the base of the interpretation of seismic anisotropy in terms of convective flow.

Coherent strain field. At large scale, the deformation due to mantle convection must be coherent over large distances in order to preserve long wavelength anisotropy. From Pn studies and models of formation of the oceanic lithosphere, it is possible to infer that anisotropy remains uniform on horizontal length-scales in excess of 1000 km. Under those assumptions it is possible to estimate the effect of LPO-induced anisotropy on body-wave speeds, and compare it with thermal effects (Montagner and Guillot 2000). The magnitude of seismic anisotropy can be estimated in a simplified geodynamical context such as an oceanic convective cell, for a simplified mineralogical composition (60% olivine, 40% orthopyroxene). The relative orientation of the crystalline populations in the kinematic field is that discussed in the previous section, and we neglect the dispersion of the orientations around this perfect or ideal orientation. The general method followed for the calculation of anisotropic parameters is summarized in Figure 2. Body-wave velocities are calculated along 4 vertical profiles, from the surface down to the 410-km-discontinuity (Fig. 3): at the mid-ocean ridge and along a slab, which are characteristic of hot or cold vertical currents; for a 7-Myr-old lithosphere and for a 30Myr-old lithosphere from the ridge, where horizontal currents are dominant. Some other profiles are reported in Montagner and Guillot (2000) for a pure olivine upper mantle. The amplitude of the temperature effect (ΔT) can be roughly estimated by comparing the ridge and the slab profiles displaying the same orientations of minerals. It is found around 3 to 4%. The differences in PH-waves velocities (horizontally-polarized P-waves) between

Seismic Anisotropy and Global Geodynamics

357

Figure 2. Different steps from mineralogical anisotropy at microscopic scale to observable seismic anisotropy at large scale.

vertical and horizontal currents are mostly due to the orientation of crystals (effect symbolized by Δα on Fig. 3). When comparing these velocities with those obtained for an equivalent isotropic system (Voigt average, see Appendix B), the magnitude of anisotropic effects on PH-wave (resp. SV-wave) velocities (Fig. 3) is about ±3%, which is slightly higher than thermal effects in vertical currents (except at the ridge). Even though anisotropy in the upper mantle should be weaker than in those experiments (natural dispersion of the crystallites), its effect on seismic properties might be as high as thermal effects; that underlines the importance of including anisotropic effects in tomographic studies. EFFECT OF ANISOTROPY ON SEISMIC WAVES The effect of seismic anisotropy on seismic waves has been extensively investigated since Love (1927). The reader is referred to the classical textbooks of Fedorov (1968) and Helbig (1994) and their application to Earth sciences (Anderson 1989, Babuska and Cara 1991). The basic theory of seismic wave propagation in anisotropic media is briefly presented in the Appendix A. The complexity of the effect of elastic anisotropy results from the general linear relationship between the stress tensor components σij and the strain tensor components εkl which involve a fourth-order elastic tensor Cijkl This relationship is named the generalized Hooke’s law:

σij = Cijkl εkl

(1)

The elastic tensor, C, in the most general case, presents 21 independent parameters. How to deal with this tensor is briefly described in the Appendix B. There are two ways to manage such complexity of C: (1) by considering either that the medium possesses symmetry properties (in order to decrease the number of independent elastic moduli) and

358

Montagner & Guillot 

Figure 3. VPH and VSV velocity profiles for a 60-40 olivine-orthopyroxene upper mantle within the convecting cell displayed in Figure 1 (from Montagner and Guillot 2000). Diamond: slab; Δ : ridge; † : 7-Myr-old plate;  : 30-Myr-old plate. The arrows represent rough approximations of the effects of orientation (Δα) and temperature (ΔT) on body wave velocities.

ultimately that it is isotropic, (2) or that the seismic anisotropy is a first-order perturbation, which permits the application of classical perturbation theories (Backus 1965). Except an isotropic medium defined by 2 elastic moduli (λ and μ), the simplest case of anisotropic medium is the transversely isotropic (TI) medium (hexagonal symmetry), characterized by 5 independent parameters A, C, F, L, N (Love 1927) when the symmetry axis is vertical. This parameterization with a vertical symmetry axis was used in deriving PREM (preliminary reference Earth model) by Dziewonski and Anderson (1981). It corresponds to the most general case for a spherically symmetric model. A slightly more complex parameterization is the TI case with a tilted symmetry axis (hexagonal symmetry), which adds 2 additional angular parameters defining the orientation of the axis. The perturbation theory is well suited for investigating the anisotropic structure of the deep Earth. As demonstrated in the first section, seismic anisotropy is not a second-order effect but, due to several averaging processes, it is usually smaller than 10%, enabling the application of perturbation theories. Seismologists are working in this framework for body waves, surface waves and normal modes. Body waves From a theoretical point of view, a first application of perturbation theory to body waves was performed by Backus (1965) and then extensively applied (Crampin 1984, for a first review paper). In the case of a weak anisotropy, we can consider that the

Seismic Anisotropy and Global Geodynamics

359

polarization of the three solutions of the Christoffel equation (see Appendix A) is very close to P, SH and SV waves. Let us call VP, VS1 and VS2 the velocities of quasi-P, quasiSV and quasi-SH waves. We draw the attention that since the sagittal plane (plane defined by the ray trajectory) is not necessarily a symmetry plane, the SV and SH components of the incoming wave may be different from the S1 and S2 components of quasi SH and SV waves. For a wave propagating horizontally in the plane (1,2) with azimuth Ψ and where 3- is the vertical direction, the different velocities are given by the following expressions in the symmetry planes: ρVP2 = 1/8 (C1111+2(C1122+C1212)+3C2222) + 1/2 (C1111–C2222)cos2ψ + (C2111+C1222)sin2ψ + 1/8 (C1111–2(C1122+C1212)+C2222)cos4ψ + 1/2 (C2111–C1222)sin4ψ 2

ρVS1 = 1/8 (C1313+C2323) + 1/2 (C1313-C2323)cos2ψ + C2313sin2ψ

(2a) (2b)

ρVS22 = 1/8 (C1111–2(C1122-C1212)+C2222) – 1/8 (C1111–2(C1122+C1212)+C2222)cos4ψ – 1/2 (C2111–C1222)sin4ψ

(2c)

These expressions are not general, but are extensively used when assuming a TI medium with any symmetry axis. This case is a very good approximation for the upper mantle pyrolitic model (Estey and Douglas 1986). For body waves, the evidence of anisotropy primarily results from the investigation of the splitting in teleseismic shear waves (Fig. 4a) such as SKS (Vinnik et al. 1984, 1989a,b, 1991, 1992; Silver and Chan 1988, 1991; Ansel and Nataf 1989), ScS (Ando 1984, Fukao 1984) and S (Ando et al. 1983; Bowman and Ando 1987; Fischer et al. 1996; Gaherty and Jordan 1995), which is corroborated by the evidence of P-wave anisotropy (Babuska et al. 1984, 1993). Among these different observations, the splitting information derived from SKS is probably the less ambiguous and has been extensively used in teleseismic anisotropy investigations. See Silver (1996) and Kendall (2000) for reviews. The drawbacks of SKS observations are, that it is almost impossible to precisely locate the depth of anisotropic regions, nor to take account for a dipping symmetry axis, and that only continental areas can be extensively investigated. The rapid variation of directions of fast velocity which can be observed in some continental regions on a short spatial scale (Vinnik et al. 1989a, Hirn et al. 1995), cannot be explained by a very deep anisotropy and the origin of anisotropy is confined to the first 410 km, either in the lithosphere or in the top of the asthenosphere. However, there are some first attempts to introduce more complex radial models of anisotropy. A close examination of SKS data makes it necessary, in several areas, to introduce at least two layers of anisotropy with different amplitudes and directions (Silver and Savage 1994, Farra and Vinnik 1994, Girardin and Farra 1998, Savage 1998, Wolfe and Silver 1998). A map summarizing the different observations of SKS splitting is presented in Figure 4b, which is an upgrade by Savage (1999) of the compilation of Silver (1996). Surface waves Surface waves are also well suited for investigating upper mantle anisotropy. Contrary to body waves, surface waves enable us to locate anisotropy at depth but, so far, its lateral resolution (several thousands of kilometers) is very poor. Two kinds of observable anisotropy have been considered. The first one results from the well-known discrepancy between Love and Rayleigh waves (Anderson 1961, McEvilly 1964), often referred as the “polarization” anisotropy (Schlue and Knopoff 1977) or the radial anisotropy. In order to remove this discrepancy, it is sufficient to consider a TI medium with a vertical symmetry axis, characterized by five elastic parameters plus density. On a global scale, Nataf et al. (1984, 1986), by the simultaneous inversion of Rayleigh and Love wave dispersion, derived the geographical distributions of S-wave radial anisotropy

360

Montagner & Guillot 

Figure 4a. Shear wave splitting. How an incoming S-wave is split when crossing an anisotropic medium. [Adapted from Silver (1996).]

Figure 4b. Shear wave splitting. Compilation of worldwide shear wave splitting measurements. [Adapted from Savage (1999).]. This dataset includes primarily SKS data but also S, ScS, SS, and PS data.

at different depths. Lévêque and Cara (1985), Cara and Lévêque (1988) used higher mode data to display radial anisotropy under the Pacific Ocean and North America down to at least 300 km. The second kind of observable anisotropy is the azimuthal anisotropy, directly derived from the azimuthal variation of phase velocity of surface waves. It was observed for the first time on Rayleigh waves by Forsyth (1975) in Nazca plate. Since these pioneering studies, global and regional models have been derived for both kinds of anisotropy (Mitchell and Yu 1980, Montagner 1985). Tanimoto and Anderson (1985) obtained a global distribution of the Rayleigh wave azimuthal anisotropy at different

Seismic Anisotropy and Global Geodynamics

361

periods. On a regional scale, several tomographic investigations reported in the eighties the existence of azimuthal anisotropy in the Indian Ocean (Montagner 1986a), Pacific Ocean (Suetsugu and Nakanishi 1987, Nishimura and Forsyth 1989) and in Africa (Hadiouche et al. 1989). The radial anisotropy (or “polarization” anisotropy) and the azimuthal anisotropy are two different manifestations of a same phenomenon, the anisotropy of the upper mantle. Montagner and Nataf (1986) derived a technique that makes it possible to simultaneously explain these two forms of seismically observable anisotropy. The principles of this technique will be only briefly described, for the most general case of anisotropy (provided that it is small). A complete description of the whole procedure can be found in Montagner (1996, 1998). Most of these measurements are based on the phase of waves. But it is now possible to use amplitude data by looking at polarization anomalies (Yu and Park 1994, Larson et al. 1998, Laske and Masters 1998, Pettersen and Maupin 2002). From a theoretical point of view, a general slight elastic anisotropy in a plane-layered medium gives rise to an azimuthal dependence of the local phase or group velocities of Love and Rayleigh waves at point (θ,φ) along the ray of the form (Smith and Dahlen 1973, 1975): V(ω,θ,φ,Ψ) – V0(ω,Ψ) = α0(ω,θ,φ) + α1(ω,θ,φ)cos2Ψ + α3(ω,θ,φ)cos4Ψ

+ α4(ω,θ,φ)sin4ψ

+ α2(ω,θ,φ)sin2Ψ

(3)

where ω is the frequency of the wave, V0(ω,Ψ) the reference velocity of the unperturbed medium, and Ψ is the azimuth along the path. The local phase velocity V(ω,θ,φ,Ψ) is used to calculate the travel time between the epicenter E and the receiver R, which is easily related to the measurement of the along-path averaged phase velocity Vd(ω): tE→R = Δ/Vd(ω) =

R

ds

∫ E V(ω ,θ,φ ,Ψ )

(4)

Montagner and Nataf (1986), following the same approach as Smith and Dahlen (1973), displayed that simple linear combinations of the elastic tensor components Cij are sufficient to describe the two seismically observable effects of anisotropy on surface waves. For the Voigt notation of the elastic constants (two indices instead of four), see Appendix B. The 0-Ψ term corresponds to the average over all azimuths and involves 5 independent parameters, A, C, F, L, N, which express the equivalent transversely isotropic medium with vertical symmetry axis. The other azimuthal terms (2-Ψ and 4-Ψ) depend on 4 groups of 2 parameters, B, G, H, respectively describing the 2-Ψ azimuthal variation of A, L, F, and E describing the 4-Ψ azimuthal variation of A and N. Therefore, the different azimuthal terms α0, α1, α2, α3, α4, depend on 13 3-dimensional parameters, which are assumed independent: • Constant term ( 0-Ψ azimuthal term: α0) A = ρVPH2 = 3/8(C11+C22) + 1/4C12 + 1/2C66 C = ρVPV2 = C33 F = 1/2(C13+C23) (5a) 2 L = ρVSV = 1/2(C44+C55) N = ρVSH2 = 1/8(C11+C22) – 1/4C12 + 1/2C66 • 2-Ψ azimuthal term: BC = 1/2(C11-C22) GC = 1/2(C55-C44) HC = 1/2(C13-C23)

362

•

Montagner & Guillot 

BS = C16 + C26 GS = C54 HS = C36 4-Ψ azimuthal term: EC = 1/8(C11+C22) – 1/4C12 – 1/2C66 ES = 1/2(C16-C26) ,

(5b)

(5c)

where indices 1 and 2 refer to horizontal coordinates (1: North; 2: East) and index 3 refers to vertical coordinate. ρ is the density, VPH, VPV are respectively horizontal and vertical propagating P-wave velocities, VSH, VSV horizontal and vertical polarized S-wave velocities. So, the different parameters present in the different azimuthal terms are simply related to elastic moduli Cij. The corresponding kernels are detailed and some of their variations at depth are plotted in Montagner and Nataf (1986). The complete description of anisotropic effects on normal modes in the spherical case can be found Mochizuki (1986) and Tanimoto (1986). In the most general case, 13 parameters are necessary to explain surface wave data (Rayleigh and Love waves) for small anisotropy, but only 4 parameters are well resolved (Montagner and Jobert 1988): the azimuthally averaged S-wave velocity VS, the radial anisotropy expressed through the ξ parameter [ξ = (VSH/VSV)2] where VSH (respectively VSV) is the velocity of S-wave propagating horizontally with horizontal transverse polarization (respectively with vertical polarization), and the G (GC, GS) parameters expressing the horizontal azimuthal variation of VSV. ξ was introduced in the reference Earth model PREM (Dziewonski and Anderson 1981) down to 220 km in order to explain a large dataset of free oscillation eigenfrequencies and body wave travel times. The other elastic parameters can be derived by using constraints from petrology in order to reduce the parameter space (Montagner and Anderson 1989a). Two extreme petrological models were used to derive the necessary correlations between anisotropic parameters, the pyrolite model (Ringwood 1975) and the piclogite model (Anderson and Bass 1984, 1986; Bass and Anderson 1984). In the depth inversion process, the smallest correlations between parameters of both models were kept and included in the a priori correlation matrix on parameters. This approach was followed by Montagner and Anderson (1989b) to derive an average reference earth model, and by Montagner and Tanimoto (1991) for the first global 3-D anisotropic model. Figure 5 shows what is expected for the observable parameters VS, ξ, G, ψ G in the case of a simple convective cell with LPO. In terms of convective flow, radial anisotropy ξ expresses its vertical (ξ < 1) or horizontal character (ξ > 1), and the azimuthal anisotropy G, can be related to the horizontal flow direction. Conversely, the three maps of VS, ξ , G, can be interpreted in terms of convective flow. These three pieces of information are necessary to correctly interpret the data. For example, upwellings or downwellings are both characterized by a weak or negative ξ parameter, but a correlative positive or negative δVS discriminates between these possibilities. By simultaneously inverting at depth for the different azimuthal terms of Rayleigh and Love waves, it is therefore possible to separate the lateral variations in temperature from those induced by the orientation of minerals. Such an interpretation should however be erroneous in waterrich mantle regions where LPO of minerals such as olivine is no simply related to the strain field (Jung and Karato 2001). Also note that this interpretation of anisotropy in terms of convective flow, using S-waves speeds only, could be erroneous in a olivinepoor mantle; see Mainprice et al. (2000), e.g., for the relations between lattice preferred orientations and VP and VS in the whole orientation space. The complete tomographic technique (regionalization + inversion at depth) has been applied for investigating either regional structures of the Indian Ocean (Montagner and

Seismic Anisotropy and Global Geodynamics

363

Jobert 1988, Debayle and Lévêque 1997), of the Atlantic Ocean (Mocquet et al. 1989, Silveira et al. 1998), of Africa (Hadiouche et al. 1989, Debayle et al. 2001), of Pacific Ocean (Nishimura and Forsyth 1989, Bussy et al. 1993), of Antarctica (Roult et al. 1994), Australia (Debayle and Kennett 2000, Simmons et al. 2002) and Central Asia (Griot et al. 1998a,b) or global structure (Montagner and Tanimoto 1990, 1991; Montagner 2002).

Figure 5. The seismic observable parameters VS, ξ, G,ψG associated with a simple convecting cell in the upper mantle, assuming lattice-preferred orientation of anisotropic minerals such as olivine. A vertical flow is characterized by a negative ξ radial anisotropy (ratio between VSH and VSV ) and a small azimuthal anisotropy (G ≈ 0). An upwelling (resp. downwelling) is characterized by a large positive (resp. negative) temperature anomaly inducing δVS < 0 (resp. δVS > 0). A predominant large-scale horizontal flow will be translated into a significant amplitude of the G azimuthal anisotropy and its orientation will reflect the direction of flow (with a 180° ambiguity).

Comparison between surface wave anisotropy and SKS splitting data It can be noted that the anisotropic parameters, linear combinations of elastic moduli Cij, derived from surface waves, also come up when you consider the propagation of body waves in symmetry planes for a slightly anisotropic medium (see section entitled Body Waves). A global investigation of anisotropy inferred from SKS body wave splitting measurements (delay times and directions of maximum velocities) has been undertaken by different authors (Vinnik et al. 1992, Silver 1996, Savage 1999). Unfortunately, most

364

Montagner & Guillot 

of SKS measurements have been done in continental parts of the Earth, and very few in oceans. It turns out that a direct comparison of body wave and surface wave datasets is now possible (Montagner et al. 2000). If the anisotropic medium is assumed to be characterized by a horizontal symmetry axis with any orientation (that is a very strong assumption), a synthetic dataset of SKS delay times and azimuths can be calculated from the global distribution of anisotropy derived from surface waves, by using the following equation: h ⎤ ρ ⎡Gc( z) Gs( z) (6) δtSKS = ∫0 dz cos2Ψ ( z) + sin 2Ψ (z )⎥ ⎢ L ⎣ L(z) L(z ) ⎦ where δtSKS is the integrated travel time for the depth range 0-h, where the anisotropic parameters Gc(z),Gs(z) and L(z) are the anisotropic parameters retrieved from surface waves at different depths. It is remarkable to realize that only the G-parameter (expressing the SV-wave azimuthal variation) is present in this equation. From Equation (6), we can infer the maximum value of delay time δtSKSmax and the corresponding azimuth ΨSKS: 2

δt

max SKS

⎧⎪ h ρ Gc( z) ⎫⎪ ⎧⎪ h ρ Gs( z) ⎫⎪ = ⎨∫0 dz ⎬ + ⎨∫0 dz ⎬ L L(z ) ⎪⎭ ⎪⎩ L L(z ) ⎪⎭ ⎪⎩ h

tan 2ΨSKS =

∫0 dz h

∫0 dz

2

(7)

ρ Gs(z ) L L(z)

ρ Gc( z)

(8)

L L(z)

However, Equation (6) is approximate and only valid when the wavelength is much larger than the thickness of layers. It is possible to make more precise calculations by using the technique derived for 2 layers by Silver and Savage (1994) or by using the general expressions given in Rumpker and Silver (1998) and Montagner et al. (2000). With Equations (7) and (8), a synthetic map of the maximum value of delay time δtSKSmax can be obtained by using a 3-D anisotropic surface wave model. A detailed comparison between synthetic SKS derived from AUM (Montagner and Tanimoto 1991) and observed SKS (Silver 1996) was presented in Montagner et al. (2000). Figure 6 shows such a map for the Pacific hemisphere, by using a new anisotropic surface wave model (Montagner 2002) derived from the data of Ekström and Dziewonski (1998). First of all, the comparison shows that both datasets are compatible in magnitude but not necessarily in directions. Some contradictions between measurements derived from surface waves and from body waves have been noted. The agreement of directions is correct in tectonically active areas but not in old cratonic zones. The discrepancy in these areas results from the rapid lateral change of directions of anisotropy at a small scale. These changes stem from the complex history of these areas, which have been built by successive collages of continental pieces. It might also result from the hypothesis of horizontal symmetry axis, which was shown to be invalid in many areas (Plomerova et al. 1996). The positive consequence of this discrepancy is that a small scale mapping of anisotropy in such areas might provide clues for understanding the processes of growth of continents and mountain building. Contrary to surface waves, SKS-waves have a good lateral resolution, and are sensitive to the short wavelength anisotropy just below the stations. But their drawback is that they have a poor vertical resolution. On the other hand, global anisotropy tomography derived from surface waves only provides long wavelength anisotropy (poor lateral resolution) but enables the location of anisotropy at depth.

Seismic Anisotropy and Global Geodynamics

365

Figure 6. Distributions of synthetic delay time δtSKSmax and azimuth ΨSKS at the surface of the Earth, such as derived from the anisotropic tomographic model of Montagner (2002), derived from data of Ekström and Dziewonski 1998). The crustal part of the 3SMAC -model (Ricard et al. 1996) has been removed. The synthetic map of SKS is calculated by using the method of Montagner et al. (2000), from the G-distribution of VSV azimuthal anisotropy. The length of lines is proportional to δtSKS.

The long wavelength anisotropy derived from surface waves will display the same direction as the short wavelength anisotropy inferred from body waves only when largescale vertical coherent processes are predominant. As demonstrated by Montagner et al. (2000), the best agreement between observed and synthetic SKS can be found when only layers in the uppermost 200 km of the mantle are taken into account. In some continental areas, short scale anisotropy, the result of a complex history, might be important and even might mask the large-scale anisotropy more related to present convective processes. This last statement settles on the following observation: when making a comparison with plate velocities directions, it is found a good statistical agreement between both directions (Fig. 7); this result shows that to first order, seismic anisotropy is reflecting the large scale plate tectonic motions. The differences between anisotropy and tectonic plate directions are related to more complex processes, as we will see below in the section Oceanic Plates. ANISOTROPY IN THE DIFFERENT LAYERS OF THE EARTH AND THEIR GEODYNAMIC APPLICATIONS

Before considering measurements of anisotropy at global, regional or local scales, let us consider the laterally averaged Earth or equivalently, the spherically symmetric reference Earth models.

366

Montagner & Guillot 

Figure 7. Histogram of the difference between plate velocities directions and synthetic SKS anisotropy azimuths in the Pacific plate.

Reference 1-D Earth models

Because tomographic models are derived from a reference model by linearized inversion schemes, there is a need for good starting radial reference models. The quality of the reference model will strongly condition the outcome of inversions for 3-D models. The most general case of anisotropy for a spherically symmetric earth is the transverse isotropy with vertical symmetry axis (also termed radial anisotropy). As discussed in the previous section, such a medium can be described using six functions of radius r, the density ρ, the wave velocities, VPH, VSV, and 3 anisotropic parameters ξ, φ and η readily related to A, C, F, L, N defined in the previous section. PREM (Dziewonski and Anderson 1981) first introduced the radial anisotropy in the uppermost 220 km of the mantle. However, some aspects of the normal mode data are not well explained by PREM (Montagner and Anderson 1989a), for example fundamental toroidal mode and body wave travel times. There were several attempts to reconcile normal mode data and body wave data (Widmer 1991, Montagner and Kennett 1996). Several robust features have been found regarding the radial anisotropy, from such inversions based on the different body-wave models [IASP91 (Kennett and Engdahl 1991); SP6 (Morelli and Dziewonski 1993); AK135 and AK303 (Kennett et al. 1996)]. First of all, anisotropy is significant in the whole upper mantle with a minimum value in the depth range 300-500 km. It is very small in the whole lower mantle except in a zone we call the “lower transition zone” (between the 660-km-discontinuity and 900-km depth)

Seismic Anisotropy and Global Geodynamics

367

and in the D′′-layer. An interesting feature of these models is the existence of a complex radial anisotropy in the transition zone (410-660 km) and at the uppermost lower mantle (660-900 km), each zone presenting a different ξ (= (VSH / VSV)2 – 1) -value pattern. The existence of anisotropy in the D′′-layer seems robust as well, but is difficult to interpret, with a small S-wave anisotropy but large P-wave and η anisotropies. Therefore, these new reference Earth models provide some indication of the existence of anisotropy in three depth ranges, the first one in D′′-layer at the core-mantle boundary, the second one around the 660-km-discontinuity. Independent seismological studies tend to corroborate these findings in the upper mantle, the transition zone, and the D′′-layer. Evidence of anisotropy in the upper 410 km of the mantle

The uppermost mantle down to 410 km is the range where the existence of seismic anisotropy is now widely recognized and well documented. Azimuthal variations have been found for body waves and surface waves in different areas of the world. During the last years, the shear wave splitting, primarily on SKS waves, was extensively used for studying continental deformation, but very few studies using body waves are devoted to oceanic areas. Conversely, global anisotropic tomographic models have been derived during the last 10 years from surface waves, but they are the most reliable below oceanic areas. Therefore the comparison of body wave and surface wave data is still in its infancy. However, as shown by Montagner et al. (2000) and Vinnik et al. (2002), such a comparison is providing encouraging results. The seismic anisotropy in the Earth can therefore be retrieved by different methods from different datasets. We will only present some examples of interesting applications of anisotropy in large scale geodynamics and tectonics. The application of seismic anisotropy to geodynamics in the upper mantle is straightforward, if we assume that fastpolarization axis of mineralogical assemblages is in the flow plane parallel to the direction of flow (Fig. 1). Seismic anisotropy in the mantle is therefore reflecting the strain field prevailing in past (frozen-in anisotropy) for shallow layers or present convective processes in deeper layers. Therefore, it makes it possible to map convection in the mantle. It must be noted that, when only the radial anisotropy is retrieved, its interpretation is non-unique. A fine layering of the mantle can also generate such a kind of anisotropy, and neglecting the azimuthal anisotropy can bias the amplitude of radial anisotropy. As mentioned in the previous section (Fig. 5), a complete interpretation of anisotropic tomography maps makes it necessary to simultaneously consider the maps of VS, ξ and G parameters in order to separate the effects of temperature and orientation of flow. A discussion for the Earth can be found in Montagner and Guillot (2000). We will only focus on geodynamic consequences for oceanic plates and continents. We will tentatively suppose that the direction of the flow in the mantle is directly related to the direction of azimuthal anisotropy or the sign of the radial anisotropy. This is approximately true in a rich-olivine mantle. A complete modeling of the relations between mantle flow and preferential orientations doesn’t exist now, but this hypothesis seems quite reasonable, especially in regions where the strain field could have been quasistationary for millions of years. Oceanic plates

Figure 8 shows 3 maps at depths of 100 km and 200 km displaying VSV velocity anomalies (Fig. 8a) and the 2 kinds of anisotropy, which can be retrieved by simultaneous inversion of Rayleigh and Love waves constant 0-Ψ and azimuthal terms of Equation (5) from the model of Montagner (2002). On Figure 8b, the equivalent radial anisotropy of the medium, for S-wave expressed through the ξ parameter, is displayed. The maps of

368

Montagner & Guillot 

Figure 8c are the distributions of the G-parameter related to the azimuthal variation of SV-wave velocity. The maximum amplitude of G is around 5% and is rapidly decreasing as depth is increasing. The distribution of anisotropy has completely different patterns and amplitudes at these 2 depths (100 and 200 km). It shows that, to the first order, the agreement of directions of maximum velocity with plate tectonics is correct in the depth range 100 to 300 km (Montagner 1994). However, the azimuth of G-parameter can largely vary as a function of depth (Montagner and Tanimoto 1991). For instance, at shallow depths (down to 60 km), the maximum velocity can be parallel to mountain belts or plate boundaries (Vinnik et al. 1991, Silver 1996, Babuska et al. 1998), and orthogonal at large depth. This means that, at a given place, the orientation of fast axis is a function of depth, making difficult the interpretation of SKS splitting. Since convective flow below oceans is dominated by large-scale plate motions, the long wavelength anisotropy found in oceanic lithospheric plates, should be similar to the smaller-scale anisotropy which should be measured from body waves. By the way, one of the first evidences of anisotropy was found in the Pacific Ocean by Hess (1964) for Pnwaves. Since that time, there were many measurements of the subcrustal anisotropy (see Babuska and Cara 1991, for a review). So far, there are very few measurements of anisotropy by SKS splitting in the oceans. Due to the lack of seismic stations on the seafloor (with the exception of H2O halfway between Hawaii and California), the only measurements available for SKS were performed in stations located on ocean islands (Ansel and Nataf 1989, Kuo and Forsyth 1992, Russo and Okal 1999, Wolfe and Silver 1998), which are by nature anomalous objects, such as volcanic hotspots where the strain field is perturbed by the ascending material and not necessarily representative of the main mantle flow field. SKS splitting was measured during the temporary MELT experiment on the East-Pacific Rise (Wolfe and Solomon 1998) but the orientation of the splitting is in disagreement with the predictions of Blackman et al. (1996). Walker et al. (2001) presented a first measurement of SKS splitting at H2O, but it is in disagreement with independent SKS splitting measurement at the same station by Vinnik et al. (2002) and with surface wave anisotropy (Montagner 2002). There were other attempts of determining anisotropy for other kinds of body waves such as ScS and multiple ScS (Ando 1984, Farra and Vinnik 1994, Gaherty et al. 1995), or differential times of sS-S, or SS-S waves (Kuo et al. 1987, Sheehan and Solomon 1991, Fischer and Yang 1994, Gaherty et al. 1995, Fouch and Fischer 1996, Yang and Fischer 1994), however, only surface wave inversion can provide a 3-D map of anisotropy below oceans. Radial cross-sections show that the ξ parameter is usually negative and small (in any case smaller than average), where flow is primarily radial (mid-ocean ridges and subducting zones). Between plate boundaries, oceans display very large areas with a large positive radial anisotropy such as in the Pacific Ocean (Ekström and Dziewonski 1997), characteristic of an overall horizontal flow field. Oceanic plates are zones where the comparison between directions of plate velocities (Minster and Jordan 1978) or NUVEL-1 (DeMets et al. 1990) and directions of G-parameter is the most successful ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Figure 8 (next page). Result of the simultaneous inversion of Rayleigh and Love waves dispersion and their azimuthal variations at 100 km depth (left) and 200 km depth (right). (adapted from Montagner 2002). (Top): Distribution of the VSV parameter in %. (Middle): ξ distributions in % with respect to ACY400 (Montagner and Anderson 1989b). Be aware that ξ anomalies are plotted with respect to a reference value different from 0. (Bottom): Anisotropy map of the G-parameter (VSV azimuthal anisotropy).

Seismic Anisotropy and Global Geodynamics

369

370

Montagner & Guillot 

in the asthenosphere down to 250-300 km (Montagner 1994). Conversely, such a comparison is more controversial below plates bearing a large proportion of continents, such as the European-Asian plate, characterized by a very small absolute motion in the hotspot coordinate system. Oceans are the areas where Plate tectonics applies almost perfectly. The large-scale azimuthal anisotropy within and below lithosphere is closely related to plate motions (Montagner 1994) and modeled in this framework (Tommasi et al. 1996). The map of the G-parameter at 100 km shows that the azimuthal anisotropy is very large along spreading ridges with a large asymmetry for the East Pacific rise. The direction of anisotropy is in very good agreement with plate motion. The anisotropy is large as well in the middle of the Pacific plate, but it can be observed that there is a line of very small azimuthal anisotropy almost parallel to the EPR. This linear area of small anisotropy was coined Low Anisotropy Channel by Montagner (2002). They are presumably related to cracking within the Pacific plate and/or to secondary convection within and below the rigid lithosphere, predicted by numerical and analog experiments. These new features provide strong constraints on the decoupling between the plate and asthenosphere. The existence and location of these LACs might be related to the current active volcanoes and hotspots (possibly plumes) in Central Pacific. LACs, which are dividing the Pacific Plate into smaller units, might indicate a future reorganization of plates with ridge migrations in the Pacific Ocean. They call for more thorough numerical modelling. Continents

Seismic anisotropy can provide fundamental information on the structure of continents, their root and the geodynamic processes involved in mountain building and collision between continents (Vinnik et al. 1992, Silver 1996) such as in Central Asia (Griot et al. 1998a,b). ξ is usually very heterogeneous below continents in the first 150200 km of depth with positive or negative areas according to geology. But it seems to display a systematic tendency of being positive at larger depth (down to 300 km), whereas it is very large in the oceanic lithosphere in the depth range 50 to 200 km and decreases rapidly at larger depths (Montagner 1994). Conversely, radial anisotropy is displaying a maximum (though smaller than in oceanic lithosphere) below very old continents (such as Siberian and Canadian Shield) in the depth range 200 to 400 km. Seismic anisotropy below continents, sometimes confined to the upper 220 km (Gaherty and Jordan 1995) can still be significant below. A more quantitative comparison of radial anisotropy between different continental provinces is presented in Babuska et al. (1998), and demonstrates systematic differences according to the tectonic context. The existence of positive large-scale radial anisotropy below continents at depth might be a good indicator of the continental root which was largely debated since the presentation of the model of tectosphere by Jordan (1978, 1981). If we assume that this maximum of anisotropy is related to an intense strain field in this depth range, it might be characteristic of the boundary between continental lithosphere and “normal” upper mantle material. And our results show that the root of continents as defined by anisotropy is located between 200 and 300 km. Such a result was recently confirmed for the Australian continent by Debayle and Kennett (2000) and Simmons et al. (2002). It must be emphasized that the anisotropy near the surface is probably different from the deep one. Part of the observed anisotropy might be related to the fossil strain field prevailing during the setting of materials and the deeper part is related to the present strain field. If we bear in mind that the anisotropy displayed from surface waves is the long wavelength filtered anisotropy (approximately 1500 km), it can be easily understood that the average anisotropy displayed from surface waves in the first 200 km might be very different from the one found from body waves. The tectonic structure of continents

Seismic Anisotropy and Global Geodynamics

371

is the result of a long and complex history. The characteristic length scale under continents is related to the size of blocks successively accreted to existing initial cores and probably smaller than 1500 km. This statement is supported by different studies of SKS body waves that demonstrate that the direction of maximum velocity can change on horizontal scales smaller than 100 km (Vinnik et al. 1991, Silver 1996). The fact that we do not observe a systematic behaviour in the first hundreds of kilometers for similar continental geological zones, does not mean that anisotropy is not present but only that its characteristic scale is different from one region to another one. Due to the low pass filtering effect of surface wave tomographic technique, the long wavelength signature of anisotropy is diluted. The measurement of seismic anisotropy can provide fundamental information for the understanding of tectonic processes. Silver (1996) and Savage (1999) present reviews of the information provided by shear-wave splitting beneath continents. The poor lateral resolution of global scale anisotropic tomography can be considered as a strong limitation in continental areas. This technique can only be efficiently applied in areas where largescale coherent processes are present. The best candidate where this condition is fulfilled in continents, is the collision zone between India and Asia, where the applicability of plate tectonics can be questioned. Griot et al. (1998a) undertook such an investigation in Central Asia. The primary goal of this study was the discrimination between two competing extreme models of deformations, the heterogenous model of Avouac and Tapponnier (1993) and the homogeneous model of England and Houseman (1986). It was necessary to use shorter wavelength surface waves (40-200 s) in order to obtain a lateral resolution of 350 km. Synthetic models of seismic anisotropy can be inferred from the heterogeneous and homogeneous models. In order to perform correct and quantitative comparisons between observed seismic anisotropy and the deformation models, the short wavelengths of the synthetic models (spatial scale smaller than 350 km) were filtered out. The statistical comparison between observed and synthetic azimuthal anisotropies for both models enables to determine in different depth ranges, which deformation model dominates. Griot et al. (1998b) show that the heterogeneous model is in better agreement with observations in the first 200 km, whereas the homogeneous model better fits the deep anisotropy below 200 km. We must be aware that such a comparison is only valid from a statistical point of view and that a comparison at a more local scale (the scale of body wave measurements) might display some differences with the observed SKS anisotropy. This kind of investigation only underlines large scale ongoing and prevailing active processes and is not devoted to a precise measurement of anisotropy at any specific place. It demonstrates that in the uppermost 200 km, the tectonics is well described by the relative movement of continental blocks and that plate tectonics can apply to these plate-like areas. Seismic anisotropy can now be used for making quantitative measurements. By using simultaneously splitting data and geodetic data, Silver and Holt (2002) were able to calculate the relative motion of lithospheric plate and asthenosphere. They show that the mantle trails the hotspot motion of the plate. This results show a weak coupling between lithosphere and asthenosphere, in contrast with the Pacific plate, where the coupling (reflected by plate direction) is the first-order effect in the uppermost 200 km, whereas LACs (Low Anisotropy channels) suggesting some decoupling, is a second-order effect. The fossil shallow anisotropy (first 100 kilometers) that reflects the past strain field, should be very useful for understanding the processes involved in surficial tectonics (see review in Savage 1999). If we were able to determine the age of establishment of this shallow anisotropy, the measurement of this kind of anisotropy should open wide a new field in Earth sciences: the Paleo-Seismology which might provide fundamental

372

Montagner & Guillot 

information to structural seology. Anisotropy in the transition zone

The transition zone plays a key role in mantle dynamics, particularly the 660-kmdiscontinuity which might inhibit the passage of matter between the upper and the lower mantle. Its seismic investigation is made difficult on global scale by the poor sensitivity of fundamental surface waves in this depth range and by the fact that teleseismic body waves recorded at continental stations from earthquakes primarily occurring along plate boundaries have their turning point below the transition zone. However, Montagner and Kennett (1996), by using eigen-frequency data, display some evidence of radial anisotropy in the upper (410-660 km) and lower (660-900 km) transition zones. Another important feature of the Transition zone is that, contrarily to the rest of the upper mantle, the upper transition zone is characterized by a large degree 2 pattern (Masters et al. 1982), and to a less extent, the degree 6. Montagner and Romanowicz (1993) explained this degree 2 pattern by the predominance of a simple large-scale flow pattern characterized by two upwellings in central Pacific Ocean and Eastern Africa and two downwellings in the Western and Eastern Pacific Ocean. This scheme was corroborated by the existence in the upper transition zone of a slight but significant radial anisotropy displayed by Montagner and Tanimoto (1991) and Roult et al. (1990). In the Transition zone, the pattern of radial anisotropy is dominated by degree 4 in agreement with the prediction of this model. Therefore, the observations of the geographical distributions of degrees 2, 4, 6 in the transition zone are coherent and spatially dependent. Montagner (1994) compared these different degrees to the corresponding degrees of the hotspot and slab distribution. In this simple framework, the distribution of plumes and slabs are merely a consequence of the large-scale simple flow in the transition zone. The existence of anisotropy close to the 660-km-discontinuity was also found by Vinnik and Montagner (1996) below Germany and Vinnik et al. (1998) in central Africa. By studying P-to-S converted waves at the GRF network and at GEOSCOPE station BNG in central Africa, they observed that part of the initial P-wave is converted into SH-wave. This signal can be observed on the transverse component of seismograms. The amplitude of this SH-wave cannot be explained by a dipping 660-km-discontinuity and it constitutes a good evidence for the existence of anisotropy just above this discontinuity. However, there is some evidence of lateral variation of anisotropy in the transition zone as found by the investigation of several subduction zones (Fischer and Yang 1994, Fischer and Wiens 1996). Fouch and Fischer (1996) present a synthesis of these different studies and show that some subduction zones such as Sakhalin Islands requires deep anisotropy in the transition zone, whereas others such as Tonga do not need any anisotropy. They conclude that their data might be reconciled by considering the upper transition zone (410-520 km) intermittently anisotropic, and the rest of the transition zone might be isotropic. The evidence of anisotropy in the transition zone was confirmed recently by 2 independent studies, using different datasets. Wookey et al. (2002) present evidence of very large S-wave splitting (up to 7 s) in the vicinity of the 660-km-discontinuity between Tonga-Kermadec subduction zone and Australia. On a global scale, Trampert and van Heijst (2002) show a robust long-wavelength azimuthal anisotropic structure in the transition zone. The rms amplitude of lateral variations of G is about 1%. The interpretation of these new exciting results is not obvious but they confirm that the transition zone might be a mid-mantle boundary layer. Anisotropy in the D′′-layer

During the last years, the structure of the mysterious D′′-layer has been extensively investigated. It is not the goal of this paper to review these different studies which make

Seismic Anisotropy and Global Geodynamics

373

use of different kinds of body waves but only to draw attention to recent results which make evident the presence of anisotropy in the D′′-layer (Vinnik et al. 1989, Lay and Young 1991). The search for anisotropic structures in the D′′-layer might provide fundamental information on boundary layer shear flow, partial melt or slab remnants at core-mantle boundary. There seems to be a general consensus that, except the D′′-layer, the lower mantle does not contribute significantly to shear wave splitting observation. However, we must be aware that it might not be valid everywhere, and part of the anisotropy attributed to D′′-layer might originate in the lowermost mantle. For example, differential splitting between phases such as P660S and SKS phases, or between SKS and SKKS, indicate that at least some regions of the lower mantle contain significant anisotropy (Iidaka and Niu 1998, Barruol and Hoffman 1999, Vinnik et al. 1997). Actually, the main drawback to investigate anisotropy in the lower mantle results from the limited crossing ray coverage. In addition, it is necessary to get rid of the influence of the anisotropy in the upper mantle. Therefore, it is not yet clear if lower mantle anisotropy is only localized in the bottom boundary layer (D′′ layer) or if there are some areas within the lower mantle with significant anisotropy. So far, there is clear evidence of anisotropy in the D′′-layer. By studying Sd-waves (S diffracted waves at the Core-Mantle boundary), Vinnik et al. (1995) found that SVdwaves is delayed relative to SHd-waves by 3s. Their observations are characteristic of a transversely isotropic medium with vertical symmetry axis with VSH > VSV. Other seismic observations such as anomalous diffraction of body waves (Maupin 1994) confirm this kind of anisotropy and its rapid lateral variation is made evident in many areas around the world (see review in Lay et al. (1998)) such as in Caribbean Sea (Kendall and Silver 1996), North America (Matzel et al. 1996, Garnero and Lay 1996), Indian Ocean (Ritsema 2000). The origin of anisotropy observed in D′′-layer is still debated (Kendall and Silver 1996, Kendall 2000) and might be due to horizontal layering in connection with paleo-subduction and/or aligned inclusions inducing different velocities for SV and SH (SPO anisotropy). On the same line, McNamara et al. (2001) showed that near cold downwellings, LPO may dominate, while near warm upwellings, SPO might be a more likely candidate. Numerical calculations of mantle convection (McNamara 2002) show that the slab deformation in the deep mantle, inducing large-strain deformation at high stresses can explain the presence of strong anisotropy in D′′-layer. NUMERICAL MODELING AND BOUNDARY LAYERS

In the previous section, we have highlighted the presence of seismic anisotropy in different parts of the Earth, using both body and surface wave data. This review of the presence of anisotropy in different layers of the Earth demonstrated that the anisotropy is a very general feature. However, it is not present in all depth ranges nor at all scales. As discussed in the first section, the observation of seismic anisotropy due to LPO at large scales requires several strong conditions, starting with the presence of anisotropic crystals up to the existence of an efficient large scale present or past strain field. Theoretical studies suggest that, when anisotropic minerals such as olivine are deformed, their crystallographic axes develop a systematic relationship to the principal axes of finite strain, and the amplitudes of the anisotropy is grossly proportional to the amplitude of the finite strain until it reaches a steady state (McKenzie 1979, Ribe 1989, Ribe and Yu 1991, Kaminski and Ribe 2001). Many numerical modelings of the convective mantle show that in a convective system, the strain field is not spatially uniform (Chastel et al. 1993). Streamlines are much more concentrated in boundary layers than in the middle of the cells. The consequence is that the amplitude of the strain field and of the seismic anisotropy resulting from the lattice preferred orientation (Tommasi et al. 2000), is very heterogeneous and the largest in boundary layers.

374

Montagner & Guillot 

For filling the gap between grain scale modeling and large scale anisotropy measurements, there is now a real need for making more quantitative comparisons between seismic anisotropy and numerical modeling. Gaboret et al. (2002) calculated the convective circulation in the mantle by converting perturbations of S-wave velocity into density perturbations. Figure 9 shows 2 cross-sections through the Pacific hemisphere and the associated flow lines derived from the tomographic model of Ekstrom and Dziewonski (1998). This kind of modeling enables to calculate the strain tensor and to test different hypotheses for the prevailing mechanisms of alignment, by comparison with seismic data. We present in Figure 10 a first attempt to compare azimuthal anisotropy of the global anisotropic model of Montagner (2002) and the maximum horizontal stretching rate. The cosine of the difference between the fast axis of VSV and the axis of horizontal extensional rate, is plotted. The best agreement is found for young ages in the Pacific ocean. More generally, the comparison is correct for oceanic plates but not so good for continental plates, where the resulting anisotropy displays a more complex distribution.

Figure 9. Two whole mantle cross-sections through the Ekstrom and Dziewonski tomographic model (1998) along great circles. Also shown in the inset maps are triangles which represent the locations of 3 Pacific hotspots. The superimposed black arrows in the cross-sections represent the mantle flow velocities predicted on the basis of the buoyancy forces derived from shear velocity anomalies. (adapted from Gaboret et al. 2002).

Seismic Anisotropy and Global Geodynamics

375

Figure 10. Comparison of azimuthal seismic anisotropy and the direction of maximum stretching predicted by the flow model presented in Figure 9. [(Adapted from Gaboret et al. (2002).]

Conversely, we can assume that the observation of mantle seismic anisotropy is the indication of a strong present-day strain field (at the exception of crust, topmost oceanic, and continental lithosphere where fossil anisotropy may be present). This strain field can be associated with boundary layers. In the previous sections, we saw that there are good evidences of the presence of seismic anisotropy in the D′′-layer, in the transition zone and in the uppermost mantle. These findings are summarized in Figure 11. The D′′-layer and the uppermost mantle have been related to boundary layers of the mantle convective system for a long time. The D′′-layer above the core-mantle boundary is characterized by a large degree of seismic heterogeneities and anisotropy with VSH larger than VSV. It might be at the same time, the graveyard of subducted slabs and the source of megaplumes. D′′- anisotropy can be related either to horizontal layering (latticeand shape- preferred orientations) of cold material or the presence of aligned inclusions owing to the presence of melt (Kendall and Silver 1996, McNamara et al. 2002). For the uppermost oceanic mantle, seismic anisotropy is present in both lithosphere and asthenosphere, and for oceans, some finite-element models are able to quantitatively relate lithospheric and asthenospheric strain to anisotropy (Tommasi et al. 1996). But the

376

Montagner & Guillot 

Figure 11. Cross-section of the Earth from the core-mantle boundary up to the surface. The hatched (respectively dotted) areas show where there are robust evidence of present (resp. fossil) anisotropy. B.L. stands for Boundary Layer.

presence of anisotropy in the transition zone (Montagner and Kennett 1996, Vinnik and Montagner 1996, Wookey et al. 2002, Trampert and van Heijst 2002) is fundamental and problematic, because it provides a new clue that the transition zone could act as a secondary boundary layer. The first evidence of anisotropy in the transition zone tends to favor the predominance of horizontal flow over vertical flow. The major consequence of this finding is that the transition zone (down to about 900-1000 km) is dividing, on average, the mantle into two convective systems, the upper mantle and the lower mantle. This general statement does not rule out the possibility that flow circulation between the upper and the lower mantle is occuring. But, it means that the exchange of matter between the upper and the lower mantle is difficult. It is too early to assess the amount of matter going through, from measurements of seismic anisotropy, but the complete mapping of seismic anisotropy in the transition zone (Trampert and van Heijst 2002) associated with numerical modeling might enable to put constraints on the flow between the upper and lower mantles. CONCLUSIONS : FROM SEISMIC ANISOTROPY TO ANISOTROPIC SEISMOLOGY

We have presented in this paper different observations of seismic anisotropy and their applications in geology and geodynamics. Seismic anisotropy is able to define continental roots, to investigate the coupling between different layers, to discriminate different geodynamic processes and competing convective models. Three boundary layers were so far detected by seismic anisotropy: the uppermost mantle, the transition zone (though new work is necessary to understand its role) and the D′′-layer. Other applications of seismic anisotropy can be easily found. For example, Montagner and Anderson (1989a) show that the different anisotropic parameters might be used for discriminating competing petrological models such as pyrolite or piclogite. Some seismologists claim that the temporal variation of anisotropy in the crust might be an efficient tool for investigating and monitoring the earthquake cycle (Crampin and Booth 1985, Crampin and Volti 1999). The temporal changes in shear-wave polarization were successfully used to show the variation of the stress regime before and after a volcanic eruption (Miller and Savage 2001). Seismic anisotropy starts to be used in seismic exploration for monitoring oil reservoirs. Therefore, to study time-dependent phenomena, seismic anisotropy turns out to be very efficient. The scientific potential of seismic

Seismic Anisotropy and Global Geodynamics

377

anisotropy is enormous and largely unexploited. In conclusion, the seismic anisotropy provides a new dimension in the investigation of processes of our dynamic Earth. ACKNOWLEDGMENTS

We are grateful to Don Anderson, Adam Dziewonski, Andrea Morelli, Lev Vinnik, Mike Gurnis, Jeroen Ritsema, Jeroen Tromp, Barbara Romanowicz, Anne Davaille, Alessandro Forte, Goran Ekström, for fruitful discussions. This is I.P.G.P. contribution # 1878. APPENDIX A. BASIC THEORY OF WAVE PROPAGATION IN ANISOTROPIC MEDIA

In any elastic medium, and in neglecting all additional terms due to prestress, gravity, inertial terms, the equation of motion is:

ρ

∂2 ui ∂t

2

=

∂ σ ∂x j ij

where ρ is the density, ui the components of the displacement, σij the components of the stress tensor. There is a linear relationship between the stress tensor and the strain tensor

ε kl =

1 (uk , l + ul,k ) , 2

such that:

σij = Cijkl εkl Let us consider ω the angular frequency of the wave, and k its propagation vector, while ν = k/k is a unit vector in the direction of k. The equations of motion simplify to:

ρ ω2 ui = kj kk Cijkl ul We look for a solution on the form of a plane wave: ⎛ ⎛ ν x ⎞⎞ ui = a i exp ⎜ iω ⎜ t − q q ⎟ ⎟ V ⎠⎠ ⎝ ⎝ where V is the phase velocity, ai the components of the polarization vector a, νk the unit vector of the wave vector ν. For a wave propagating in direction e1:

ρV2a1 = C1111 a1+ C1112 a2+ C1113 a3 ρV2a2 = C2111 a1+ C2112 a2+ C2113 a3 ρV2a3 = C3111 a1+ C3112 a2+ C3113 a3 These three equations can be written in a condensed way: ( T – ρ V2 I ) a = 0 where

378

Montagner & Guillot  ⎛ C1111 C1112 C1113 ⎞ T = ⎜⎜ C2111 C2112 C2113 ⎟⎟ ⎟ ⎜ ⎝ C3111 C3112 C3113 ⎠

is the Christoffel matrix, I the identity matrix. That is an eigenvalue problem which has 3 real positive roots for ρV2 with orthogonal eigenvectors. Therefore, these equations demonstrate that there are 3 body waves in every direction of phase propagation with orthogonal particle motion and with velocities which in general are different and vary with direction. In case of isotropic medium the elastic tensor is invariant with rotation, and T is a diagonal matrix: ⎛ λ + 2μ T =⎜ 0 ⎜ ⎜ ⎝ 0

0

μ 0

0⎞ 0⎟ ⎟ μ⎟⎠

The well-known isotropic velocities can be written immediately: V = VP =

λ + 2μ , ρ

and VP =

μ (repeated root). ρ

In case of weak anisotropy, we can define quasi P-waves and quasi S-waves, but in case of very strong anisotropy, the eigenvectors can have any orientation which can be very different from quasi P- and quasi S-waves. An important consequence is that the propagation of energy of a plane wave is no longer in the direction of the propagation vector. APPENDIX B. TENSORS AND MATRICES MANIPULATIONS

The effect of intrinsic anisotropy has been extensively studied in laboratory and in the field. Most crystals present in the Earth are strongly anisotropic (olivine, pyroxenes) except garnet. A medium is perfectly elastic when a linear relationship (Hooke’s law) relates the stress tensor σij and the strain tensor εkl. A general elastic medium is anisotropic since its physical properties are dependent on its geometrical orientation. The anisotropic properties of any elastic medium are usually described by its fourth-order elastic tensor Cijkl which linearly relates the stress tensor σ and the strain tensor ε, as seen in the previous section. Contraction of indices. Instead of keeping 4 indices, it is common, for sake of simplicity to write the elastic tensor Cijkl as a matrix (6 × 6) by using the following rules (the reader must be aware that for avoiding mathematical errors, it is easier to perform calculations by using the fourth-order elastic tensor Cijkl):

cijkl (3×3×3×3)

Cpq (6×6)

p = i=j if i = j; q = k = l if k = l p = 9-i-j if i≠j; q = 9-k-l if k≠ l This is the Voigt notation. It allows to obtain simplified matrix representations for elastic tensors:

Seismic Anisotropy and Global Geodynamics

•

Isotropic elastic tensor (2 independent parameters): ⎛ λ + 2μ λ ⎜ 0 λ + 2μ ⎜ 0 ⎜ 0 ⎜ 0 ⎜ 0 ⎜ 0 0 ⎜ 0 ⎝ 0

•

λ λ λ + 2μ 0 0 0

0⎞ 0⎟ ⎟ 0⎟ ⎟ μ 0⎟ 0 μ 0⎟ ⎟ 0 0 μ⎠ 0 0 0

0 0 0 0

Elastic tensor of a transversely isotropic medium with a vertical symmetry axis (5 independent parameters): ⎛ A ⎜ A − 2N ⎜ ⎜ F ⎜ ⎜ 0 ⎜ 0 ⎜ ⎝ 0

•

379

A − 2N A F 0 0 0

F F C 0 0 0

0 0 0 L 0 0

0 0 0 0 L 0

0⎞ 0⎟ ⎟ 0⎟ ⎟ 0⎟ 0⎟ ⎟ N⎠

Elastic tensor of a general orthorhombic medium with three mutually perpendicular symmetry planes (9 independent parameters): ⎛a ⎜b ⎜ ⎜c ⎜ ⎜0 ⎜0 ⎜ ⎝0

b d e 0 0 0

c e f 0 0 0

0 0 0 g 0 0

0 0 0 0 h 0

0⎞ 0⎟ ⎟ 0⎟ ⎟ 0⎟ 0⎟ ⎟ i⎠

These expressions are valid, for a given symmetry, only in a specific co-ordinate system, the crystal reference frame, where the reference axes are simply related to the symmetry elements of the medium (see Nye 1957, e.g.). In the general case, the components of the tensor are expressed in another reference frame, which can be related to the crystal reference frame by a 3-D rotation. This rotation is a linear transformation of the basis vectors of the crystal reference frame, and as any linear transformation it can be represented by a matrix A, whose components are written Ai j. The components C’mnpq of the fourth-order elastic tensor in the new reference frame, after rotation, are related to the crystal frame components Cijkl as: Cmnpq ′ = Ami Anj A kp Aql Cijkl

Note that this expression is only valid for the covariant components of the elastic tensor, in the general case. For the contravariant and mixed components, e.g., see Hladik (1995). In this paper, the elastic tensor is expressed in the usual 3-D (real) Cartesian reference frame, A represents a rotation and is then an orthogonal real matrix, so the expression above is valid for all types of components. Elastic matrix: scale of aggregates. For polycrystals, problems of averaging arise. There are different ways to make averages. Let nm be the number of different crystalline

380

Montagner & Guillot 

populations, and αq their proportion in the aggregate. Each mineral of a given population has its elasticity described by the elastic tensor C, with components Cijklq. These quantities are called stiffnesses. We also introduce the compliance tensor S, whose components Sijkl are defined by: εij = Sijkl σkl. Different averages among the most commonly used are (Watt et al. 1976): • Voigt average: hypothesis of constant deformation. We obtain for the stiffnesses: nm

q CVijkl = ∑ α q Cijkl q =1

• Reuss average: hypothesis of constant stress. We obtain for the compliances: nm

q SVijkl = ∑ α q S ijkl q =1

After some manipulations (matrix inversion in the most general case), the associated R stiffnesses Cijkl can be calculated (Nye 1957). • Voigt-Reuss-Hill average: it is only the arithmetic mean of the Voigt and Reuss averages, and has no theoretical justification. • Hashin-Shtrikman: Minimization of the elastic energy (no analytical solution in the most general case).

A more thorough description of these averages as well as others can be found in Mainprice et al. (2000). REFERENCES Aki K, Kaminuma K (1963) Phase velocity of Love waves in Japan (part 1): Love waves from the Aleutian shock of March 1957. Bull Earthquake Res Inst 41:243-259 Anderson DL (1961) Elastic wave propagation in layered anisotropic media. J Geophys Res 66:2953-2963 Anderson DL (1989) Theory of the Earth. Blackwell Scientific Publications, Oxford, UK Anderson DL, Bass JD (1984) Mineralogy and composition of the upper mantle. Geophys Res Lett 1: 637-640 Anderson DL, Bass JD (1986) Transition region of the Earth's upper mantle. Nature 320:321-328 Anderson DL, Regan J (1983) Upper mantle anisotropy and the oceanic lithosphere. Geophys Res Lett 10:841-844 Ando M (1984) ScS polarization anisotropy around the Pacific Ocean. J Phys Earth 32:179-196 Ando M, Ishikawa Y, Yamazaki F (1983) Shear wave polarization anisotropy in the upper mantle beneath Honshu, Japan. J Geophys Res 88:5850-5864 Ansel V, Nataf HC (1989) Anisotropy beneath 9 stations of the Geoscope broadband network as deduced from shear wave splitting. Geophys Res Lett 16:409-412 Avouac JP, Tapponnier P (1993) Kinematic model of active deformation in central Asia. Geophys Res Lett 20:895-898 Babuska V, Cara M (1991) Seismic Anisotropy in the Earth. Kluwer Academic Press, Dordrecht, The Netherlands Babuska V, Montagner JP, Plomerova J, Girardin N (1998) Age-dependent large-scale fabric of the mantle lithosphere as derived from surface-wave velocity anisotropy. Pure Appl Geophys 151:257-280 Backus GE (1962) Long wave elastic anisotropy produced by horizontal layering. J Geophys Res 67: 4427- 4440 Backus GE (1965) Possible forms of seismic anisotropy of the upper mantle under oceans. J Geophys Res 70:3249-3439 Barruol G, Hoffman R (1999) Upper mantle anisotropy beneath the GEOSCOPE stations. J Geophys Res 104:10757-10774 Blackman DK, Kendall JM, Dawson PR, Wenk HR, Boyce D, Morgan JP (1996) Teleseismic imaging of subaxial flow at mid-ocean ridges: Travel-time effects of anisotropic mineral texture in the mantle. Geophys J Intl 127:415-426

Seismic Anisotropy and Global Geodynamics

381

Bostock MG (1997) Anisotropic upper-mantle stratigraphy and architecture of the Slave craton. Nature 390:392-395 Bowman JR, Ando M (1987) Shear-wave splitting in the upper mantle wedge above the Tonga subduction zone. Geophys J R Astron Soc 88:25-41 Cara M, Leveque JJ (1988) Anisotropy of the asthenosphere: The higher mode data of the Pacific revisited. Geophys Res Lett 15:205-208 Chastel YB, Dawson PR, Wenk HR, Bennett K (1993) Anisotropic convection with implications for the upper mantle. J Geophys Res 98:17757-17771 Christensen NI, Lundquist S (1982) Pyroxene orientation within the upper mantle. Bull Geol Soc Am 93:279-288 Crampin S (1984) An introduction to wave propagation in anisotropic media. Geophys J R Astron Soc 76:17-28 Crampin S, Booth DC (1985) Shear-wave polarizations near the North Anatolian fault, II: Interpretation in terms of crack-induced anisotropy. Geophys J R Astron Soc 83:75-92 Crampin S, Volti T, Stefansson R (1999) A successfully stress-forecast earthquake. Geophys J Intl 138: F1-F5 Debayle E (1999) SV-wave azimuthal anisotropy in the Australian upper mantle: preliminary results from automated Rayleigh waveform inversion. Geophys J Intl 137:747-754 Debayle E, Lévêque JJ (1997) Upper mantle heterogeneities in the Indian Ocean from waveform inversion. Geophys Res Lett 24:245-248 Debayle E, Kennett BLN (2000) Anisotropy in the Australasian upper mantle from Love and Rayleigh waveform inversion. Earth Planet Sci Lett 184:339-351 Dziewonski AM, Anderson DL (1981) Preliminary Reference Earth Model. Phys Earth Planet Inter 25: 297-356 Ekström G, Dziewonski AM (1998) The unique anisotropy of the Pacific upper mantle. Nature 394:168-172 England P, Houseman G (1986) Finite strain calculations of continental deformation, 2. Comparison with the India-Asia collision zone. J Geophys Res 91:3664-3676 Estey LH, Douglas BJ (1986) Upper mantle anisotropy: A preliminary model. J Geophys Res 91:1139311406 Farra V, Vinnik LP (1994) Shear-wave splitting in the mantle of the Pacific. Geophys J Intl 119:195-218 Fedorov F (1968) Theory of Elastic waves in crystals. Plenum Press, New York Fischer KM, Wiens DA (1996) The depth distribution of mantle anisotropy beneath the Tonga subduction zone. Earth Planet Sci Lett 142:253-260 Fischer KM, Yang X (1994) Anisotropy in Kuril-Kamtchatka subduction zone structure. Geophys Res Lett 21:5-8 Forsyth DW (1975) The early structural evolution and anisotropy of the oceanic upper mantle. Geophys J R Astron Soc 43:103-162 Forsyth DW, Webb SC, Dorman LM, Shen Y (1998) Phase velocities of Rayleigh waves in MELT experiment on the East Pacific Rise. Science 280:1235-1238 Fouch MJ, Fischer KM (1996) Mantle anisotropy beneath northwest Pacific subduction zones. J Geophys Res 101:15987-16002 Fouch MJ, Fischer KM, Parmentier EM, Wysession ME, Clarke TJ (2000) Shear wave splitting, continental keels, patterns of mantle flow. J Geophys Res 105:6255-6275 Fouch MJ, Fischer KM, Wysession ME (2001) Lowermost mantle anisotropy beneath the Pacific: Imaging the source of the Hawaiian plume. Earth Planet Sci Lett 190:167-180 Fukao Y (1984) Evidence from core- reflected shear waves for anisotropy in the Earth's mantle. Nature 309:695-698 Gaboret C, Forte A, Montagner JP (2002) The unique dynamics of the Pacific hemisphere mantle and its signature on seismic anisotropy. Earth Planet Sci Lett (submitted) Gaherty JB, Jordan TH (1995) Lehmann discontinuity as the base of the anisotropic layer beneath continents. Science 268:1468-1471 Garnero EJ, Lay T (1996) Lateral variation in lowermost mantle shear wave anisotropy beneath the North Pacific and Alaska. J Geophys Res 102:8121-8135 Griot DA, Montagner JP, Tapponnier P (1998a) Surface wave phase velocity and azimuthal anisotropy in Central Asia. J Geophys Res 103:21215-21232 Griot DA, Montagner JP, Tapponnier P (1998b) Heterogeneous versus homogeneous strain in Central Asia. Geophys Res Lett 25:1447-1450 Hadiouche O, Jobert N, Montagner JP (1989) Anisotropy of the African continent inferred from surface waves. Phys Earth Planet Inter 58:61-81 Hager B, O'Connell R (1979) Kinematic models of large-scale flow in the Earth's mantle. J Geophys Res 84:1031-1048

382

Montagner & Guillot 

Helbig K (1994) Foundations of Anisotropy for Exploration Seismics. Pergamon/Elsevier Science, Oxford, UK Hess H (1964) Seismic anisotropy of the uppermost mantle under the oceans. Nature 203:629-631 Hladik J (1995) Le calcul tensoriel en physique. Paris, Masson, France. Hirn A, Jiang M, Sapin M, Diaz J, Nercessian A, Lu QT, Lépine JC, Shi DN, Sachpazi M, Pandy MR, Ma K, Gallart J (1995) Seismic anisotropy as an indicator of mantle flow beneath the Himalayas and Tibet. Nature 375:571-574 Iidaka T, Niu F (1998) Evidence for an anisotropic lower mantle beneath eastern Asia: comparison of shear wave splitting data of SKS and P660s. Geophys Res Lett 25:675-678 Jordan TH (1978) Composition and development of the continental tectosphere. Nature 274:544-548 Jordan TH (1981) Continents as a chemical boundary layer. Phil Trans R Soc London Ser A 301:359-373 Jung HY, Karato SI (2001) Water-induced fabric transitions in olivine. Science 293:1460-1462 Kaminski E, Ribe NM (2001) A kinematic model for recrystallization and texture development in olivine polycrystals. Earth Planet Sci Lett 189:253-267 Karato S-i (1989) Seismic anisotropy: mechanisms and tectonic implications. In Rheology of Solids and of the Earth. S-i Karato, M Toriumi (eds) Oxford University Press, Oxford, p 393-42 Karato SI, Li P (1993) Diffusive creep in perovskite: Implications for the rheology of the lower mantle. Science 255:771-778 Kendall JM (2000) Seismic anisotropy in the boundary layers of the mantle. In Earth's Deep Interior: Mineral physics and tomography from the atomic to the global scale. Karato S-i, Forte A, Liebermann RC, Masters G, Stixrude L (eds) Am Geophys Union Monograph, Washington, DC Kendall JM, Silver PG (1996) Constraints from seismic anisotropy on the nature of the lowermost mantle. Nature 381:409-412 Kennett BLN, Engdahl ER (1991) Traveltimes for global earthquake location and phase identification. Geophys J Intl 105:429-465 Larson EWF, Tromp J, Ekstrom G (1998) Effects of slight anisotropy on surface waves. Geophys J Intl 132:654-666 Laske G, Masters G (1998) Surface-wave polarization data and global anisotropic structure. Geophys J Intl 132:508-520 Lavé J, Avouac JP, Lacassin R, Tapponnier P, Montagner JP (1996) Seismic anisotropy beneath Tibet: Evidence for eastward extrusion of the Tibetan lithosphere? J Geophys Res 140:83-96 Lay T, Williams Q, Garnero EJ, Kellog L, Wysession ME (1998) Seismic wave anisotropy in the D'' region and its implications. In Core-Mantle Boundary Region. Gurnis M, Wysession ME, Knittle E, Buffett B (eds) Geodynamic Series 28:229-318 Lévêque JJ, Cara M (1985) Inversion of multimode surface wave data: evidence for sub--lithospheric anisotropy. Geophys J R Astron Soc 83:753-773 Lévêque JJ, Cara M, Rouland D (1991) Waveform inversion of surface-wave data: a new tool for systematic investigation of upper mantle structures. Geophys J Intl 104:565-581 Levshin A, Ratnikova L (1984) Apparent anisotropy in inhomogeneous media. Geophys J R Astron Soc 76: 65-69 Love AEH (1927) A Treatise on the Theory of Elasticity, 4th Edition. Cambridge University Press, Cambridge, UK Mainprice DG, Barruol G, Ben Ismail W (2000) The seismic anisotropy of the Earth's. mantle: From single crystal to Polycrystal. In Earth's Deep Interior: Mineral Physics and Tomography From the Atomic Scale to the Global Scale. Geophys Monogr 117:237-264 Masters G, Jordan TH, Silver PG, Gilbert F (1982) Aspherical Earth structure from fundamental spheroidalmode data. Nature 298:609-613 Matzel E, Sen MK, Grand SP (1996) Evidence for anisotropy in the deep mantle beneath Alaska. Geophys Res Lett 23:2417-2420 Maupin V (1994) On the possibility of anisotropy in the D'' layer as inferred from the polarization of diffracted S-waves. Phys Earth Planet Inter 87:1-32 McEvilly TV (1964) Central U.S. crust-upper mantle structure from Love and Rayleigh wave phase velocity inversion. Bull Seism Soc Am 54:1997-2015 McKenzie D (1979) Finite deformation during fluid flow. Geophys J R Astron Soc 58:687-715 McNamara AK, Karato SI, van Keken PE (2001) Localization of dislocation creep in the lower mantle; implications for the origin of seismic anisotropy. Earth Planet Sci Lett 191:85-99 McNamara AK, van Keken PE, Karato SI (2002) Development of anisotropic structure in the Earth's lower mantle by solid-state convection. Nature 416:310-314 Meade C, Silver PG, Kaneshima S (1995) Laboratory and seismological observations of lower mantle anisotropy. Geophys Res Lett 22:1293-1296 Miller V, Savage S (2001) Changes in seismic anisotropy after volcanic eruptions: Evidence from Mount Ruapehu. Science 293:2231-2233

Seismic Anisotropy and Global Geodynamics

383

Minster JB, Jordan TH (1978) Present-day plate motions. J Geophys Res 83:5331-5354 Mitchell BJ, Yu GK (1980) Surface wave dispersion, regionalized velocity models and anisotropy of the Pacific crust and upper mantle. Geophys J R Astron Soc 63:497-514 Mochizuki E (1986) The free oscillations of an anisotropic and heterogeneous Earth. Geophys J R Astron Soc 86:167-176 Montagner JP (1985) Seismic anisotropy of the Pacific Ocean inferred from long-period surface wave dispersion. Phys Earth Planet Inter 38:28-50 Montagner JP (1986a) First results on the three dimensional structure of the Indian Ocean inferred from long period surface waves. Geophys Res Lett 13:315-318 Montagner JP (1986b) Regional three-dimensional structures using long-period surface waves. Ann Geophys 4:B3:283-294 Montagner JP (1994) What can seismology tell us about mantle convection? Rev Geophys 32:2:115-137 Montagner JP (1998) Where can seismic anisotopy be detected in the Earth's mantle? In boundary layers. Pure Appl Geophys 151:223-256 Montagner JP (2002) Upper mantle low anisotropy channels below the Pacific Plate. Earth Planet Sci Lett (in press) Montagner JP, Anderson DL (1989a) Constraints on elastic combinations inferred from petrological models. Phys Earth Planet Inter 54:82-105 Montagner JP, Anderson DL (1989b) Constrained reference mantle model. Phys Earth Planet Inter 58: 205-227 Montagner JP, Griot DA, Lavé J (2000) How to relate body wave and surface wave anisotropies? J Geophys Res 105:19015-19027 Montagner JP, Guillot L (2000) Seismic anisotropy tomography. In Problems in Geophysics for the Next Millennium. E Boschi, G Ekström, A Morelli (eds) Editrice Compositori, Bologna, Italy, p 217-254 Montagner JP, Jobert N (1988) Vectorial Tomography. II: Application to the Indian Ocean. Geophys J R Astron Soc 94:309-344 Montagner JP, Kennett BLN (1996) How to reconcile body-wave and normal-mode reference Earth models? Geophys J Intl 125:229-248 Montagner JP, Lognonné P, Beauduin R, Roult G, Karczewski JF, Stutzmann E (1998) Towards multiscale and multiparameter networks for the next century: The French efforts. Phys Earth Planet Inter 108: 155-174 Montagner JP, Nataf HC (1986) On the inversion of the azimuthal anisotropy of surface waves. J Geophys Res 91:511-520 Montagner JP, Nataf HC (1988) Vectorial Tomography. I: Theory. Geophys J R Astron Soc 94:295-307 Montagner JP, Tanimoto T (1990) Global anisotropy in the upper mantle inferred from the regionalization of phase velocities. J Geophys Res 95:4797-4819 Montagner JP, Tanimoto T (1991) Global upper mantle tomography of seismic velocities and anisotropies. J Geophys Res 96:20337-20351 Montagner JP, Romanowicz B, Karczewski JF (1994) A first step towards an Oceanic Geophysical Observatory. EOS, Trans Am Geophys Union 75:150-154 Montagner JP, Romanowicz B (1993) Degrees 2, 4, 6 inferred from seismic tomography. Geophys Res Lett 20:631-634 Morelli A, Dziewonski AM, Woodhouse JH (1986) Anisotropy of the inner core inferred PKIKP travel times. Geophys Res Lett 13:1545-1548 Morelli A, Dziewonski AM (1993) Body wave traveltimes and a spherically symmetric P- and S-wave velocity model. Geophys J Intl 112:178-194 Nataf HC, Nakanishi I, Anderson DL (1984) Anisotropy and shear velocity heterogeneities in the upper mantle. Geophys Res Lett 11:109-112 Nataf HC, Nakanishi I, Anderson DL (1986) Measurement of mantle wave velocities and inversion for lateral heterogeneity and anisotropy, III. Inversion. J Geophys Res 91:7261-7307 Nicolas A (1993) Why fast polarization directions of SKS seismic waves are parallel to mountain belts? Phys Earth Planet Inter 78:337-342 Nicolas A, Boudier F, Boullier AM (1973) Mechanisms of flow in naturally and experimentally deformed peridototes. Am J Sci 273:853-876 Nicolas A, Christensen NI (1987) Formation of anisotropy in upper mantle peridotites: A review. In Composition, Structure and Dynamics of the Lithosphere/Asthenosphere System. Fuchs K, Froidevaux C (eds) American Geophysical Union, Washington, DC, p 111-123 Nishimura CE, Forsyth DW (1989) The anisotropic structure of the upper mantle in the Pacific. Geophys J 96:203-229 Nye JF (1985) Physical Properties of Crystals: Their Representation by Tensors and Matrices. Reprint (with corrections). Clarendon Press, Oxford, UK Park J, Levin V (2002) Seismic anisotropy: tracing plate dynamics in the mantle. Science 296:485-489

384

Montagner & Guillot 

Peselnick L, Nicolas A, Stevenson PR (1974) Velocity anisotropy in a mantle peridotite from Ivrea zone: Application to uper mantle anisotropy. J Geophys Res 79:1175-1182 Peselnick L, Nicolas A (1978) Seismic anisotropy in an ophiolite peridotite: Application to oceanic upper mantle. J Geophys Res 83:1227-1235 Pettersen O, Maupin V (2002) Lithospheric anisotropy on the Kerguelen hotspot track inferred from Rayleigh wave polarisation anomalies. Geophys J Intl 149:225-246 Ribe NM (1989) Seismic anisotropy and mantle flow. J Geophys Res 94:4213-4223 Ribe NM, Yu Y (1991) A theory for plastic deformation and textural evolution of olivine polycrystals. J Geophys Res 96:8325-8335 Ricard Y, Nataf HC, Montagner JP (1996) The 3S-Mac model: confrontation with seismic data. J Geophys Res 101:8457-8472 Ringwood AE (1975) Composition and Petrology of the Earth's Mantle. McGraw-Hill, New York, 618 p Ritsema J (2000) Evidence for shear wave anisotropy in the lowermost mantle beneath the Indian Ocean. Geophys Res Lett 27:1041-1044 Roult G, Rouland D, Montagner JP (1994) Antarctica II: Upper mantle structure from velocity and anisotropy. Phys Earth Planet Inter 84:33-57 Rümpker G, Silver PG (1998) Apparent shear-wave splitting in the presence of vertically varying anisotropy. Geophys J Intl 135:790-800 Russo R, Silver PG (1994) Trench-parallel flow beneath the Nazca plate from seismic anisotropy. Science 263:1105-1111 Russo RM, Okal EA (1999) Shear wave splitting and upper mantle deformation in French Polynesia: Evidence for small-scale heterogeneity related to the Society hotspot. J Geophys Res 103:15089-15107 Savage MK (1999) Seismic anisotropy and mantle deformation: What have we learned from shear wave splitting? Rev Geophys 37:65-106 Schlue JW, Knopoff L (1977) Shear--wave polarization anisotropy in the Pacific Ocean. Geophys J R Astron Soc 49:145-165 Silver PG (1996) Seismic anisotropy beneath the continents: Probing the depths of geology. Ann Rev Earth Planet Sci 24:385-432 Silver PG, Chan WW (1988) Implications for continental structure and evolution from seismic anisotropy. Nature 335:34-39 Silver PG, Chan WW (1991) Shear wave splitting and subcontinental mantle deformation. J Geophys Res 96:16429-16454 Silver PG, Holt WE (2002) The mantle flow field beneath North America. Science 295:1054-1057 Silveira G, Stutzmann E, Montagner JP (1998) Mendes-Victor L. Anisotropic tomography of the Atlantic Ocean from Rayleigh surface waves. Phys Earth Planet Inter 106:259-275 Simmons FJ, van der Hilst R, Montagner JP, Zielhuis A (2002) Multimode Rayleigh wave inversion for shear wave speed heterogeneity and azimuthal anisotropy of the Australian upper mantle. Geophys J Intl (in press) Singh S, Taylor M, Montagner JP (2002) On the presence of liquid in Earth’s inner core. Science 287: 2471-2474 Smith ML, Dahlen FA (1973) The azimuthal dependence of Love and Rayleigh wave propagation in a slightly anisotropic medium. J Geophys Res 78:3321-3333 Smith ML, Dahlen FA (1975) Correction to ‘The azimuthal dependence of Love and Rayleigh wave propagation in a slightly anisotropic medium’. J Geophys Res 80:1923 Suetsugu D, Nakanishi I (1987) Regional and azimuthal dependence of phase velocities of mantle Rayleigh waves in the Pacific Ocean. Phys Earth Planet Inter 47:230-245 Tanimoto T (1986) Free oscillations in a slightly anisotropic Earth. Geophys J R Astron Soc 87:493-517 Tanimoto T, Anderson DL (1985) Lateral heterogeneity and azimuthal anisotropy of the upper mantle: Love and Rayleigh waves 100-250s. J Geophys Res 90:1842-1858 Tommasi A, Mainprice D, Canova G, Chastel Y (2000) Viscoplastic sel-consistent and equilibrium-based modeling of oilivine preferred orientations: Implications for the upper mantle anisotropy. J Geophys Res 105:7893-7908 Tommasi A, Vauchez A, Russo R (1996) Seismic anisotropy in ocean basins: Resistive drag of the sublithospheric mantle? Geophys Res Lett 23:2991-2994 Trampert J, Woodhouse JH (1996) Global phase velocity maps of Love and Rayleigh waves between 40 and 150 seconds. Geophys J Intl 212:675-690 Trampert J, van Heijst HJ (2002) Global anisotropy azimythal anisotropy in the transition zone. Science 296:1297-1299 van der Hilst RD, Karason H (1999) Compositional heterogeneity in the bottom 1000 km of the Earth's mantle: Toward a hybrid convection model. Science 283:1885-1888 Vauchez A, Nicolas A (1991) Mountain building: strike-parallel motion and mantle anisotropy. Tectonophysics 185:183-191

Seismic Anisotropy and Global Geodynamics

385

Vinnik LP, Chevrot S, Montagner JP (1998) Seismic evidence of flow at the base of the upper mantle. Geophys Res Lett 25:1995-1998 Vinnik LP, Farra V, Romanowicz B (1989b) Azimuthal anisotropy in the earth from observations of SKS at GEOSCOPE and NARS broadband stations. Bull Seism Soc Am 79:1542-1558 Vinnik LP, Kosarev GL, Makeyeva LI (1984) Anisotropiya litosfery po nablyudeniyam voln SKS and SKKS. Dokl Akad Nauk USSR 278:1335-1339 Vinnik LP, Kind R, Kosarev GL, Makeyeva LI (1989a) Azimuthal Anisotropy in the lithosphere from observations of long-period S-waves. Geophys J Intl 99:549-559 Vinnik LP, Makayeva LI, Milev A, Usenko AY (1992) Global patterns of azimuthal anisotropy and deformations in the continental mantle. Geophys J Intl 111:433-447 Vinnik LP, Montagner JP (1996) Shear wave splitting in the mantle from Ps phases. Geophys Res Lett 23:2449-2452 Vinnik LP, Romanowicz B, Le Stunff Y, Makayeva L (1995) Seismic anisotropy in D''-layer. Geophys Res Lett 22:1657-1660 Vinnik LP, Montagner JP, Girardin N, Dricker I (2002) Saul, Shear wave splitting at H2O: A comment. Geophys Res Lett (submitted) Walker KT, Bokelmann GH, Klemperer SL (2001) Shear-wave splitting to test mantle Deformation models around Hawaii. Geophys Res Lett 28:4319-4322 Watt JP, Davies GF, O’Connell (1976) The elastic properties of composite materials. Rev Geophys Space Phys 14:541-565 Widmer R, Masters G, Gilbert F (1993) Spherically symmetric attenuation within the Earth from normal mode data. Geoph J Intl 104:541-553 Wolfe JW, Silver PG (1998) Seismic anisotropy of oceanic upper mantle: Shear wave splitting methodologies and observations. J Geophys Res 103:749-771 Woodhouse JH, Giardini D, Li XD (1986) Evidence for inner core anisotropy from free oscillations. Geophys Res Lett 13:1549-1552 Wookey J, Kendall JM, Barruol G (2002) Mid-mantle deformation inferred from seismic anisotropy. Nature 415:777-780 Yu Y, Park J (1993) Anisotropy and coupled long-period surface waves. Geophys J Intl 114:473-489 Zhang S, Karato S-i (1995) Lattice preferred orientation of olivines aggregates deformed in simple shear. Nature 375:774-777

13 Theoretical Analysis of Shear Localization in the Lithosphere David Bercovici and Shun-ichiro Karato Department of Geology and Geophysics Yale University New Haven, Connecticut 06511

INTRODUCTION Deformation in the Earth is rarely homogeneous and often occurs in narrow regions of concentrated strain referred to as zones of shear localization. Mylonites found in the continental crust are the manifestation of shear localization (White et al. 1980) and, at the largest scale, shear localization is proposed to be crucial for the generation of tectonic plates from a convecting mantle (Bercovici 1993, 1995b,a, 1996, 1998; Bercovici et al. 2000; Trompert and Hansen 1998; Tackley 1998, 2000a,b,c,d). Shear localization is apparent in many field of physics, including, for example, metallurgy (Lemonds and Needleman 1986), rock mechanics (Poirier 1980; Jin et al. 1998), granular dynamics (Scott 1996; G´eminard et al. 1999, e.g.) and glaciology (Yuen and Schubert 1979). However, basic solid-state rheologies, such as elasticity, visco-elasticity, viscous flow and even steady-state non-Newtonian viscous flow, are insufficient by themselves to generate shear-localization; this is because in all such rheologies an increase in deformation or rate of deformation results in greater resistance (i.e., stress) instead of self-weakening and loss of strength. Shear-localization tends to require dynamic feedback mechanisms wherein self-weakening is controlled by a macroscopic variable (such as temperature) or microscopic structure (grain size or microcrack density) whose evolution and concentration are themselves determined by deformation; in this way deformation can induce weakening, which subsequently causes deformation to concentrate on the weak zone (being most easily deformed), causing further weakening, and thus more focusing of deformation, and so on. One of the most fundamental manifestations of such feedback mechanisms arises from the coupling of viscous heating and temperature-dependent viscosity wherein the zone of dissipative heating weakens and thus focuses deformation, leading to further heating and weakening; this mechanism is thought applicable to problems in metal weakening, glacial surges, and lithosphere dynamics (Schubert and Turcotte 1972; Yuen and Schubert 1979; Poirier 1980; Balachandar et al. 1995; Bercovici 1996; Thatcher and England 1998). In granular media, localization is due to dilation of the medium leading to effectively weaker rarifed zones that concentrate deformation, which in turn agitates the medium causing further dilation (G´eminard et al. 1999); this phenomenon is of potential importance in earthquake dynamics (Scott 1996; Marone 1998; Segall and Rice 1995; Sleep 1995, 1997; Mora and Place 1998). Another shear localization mechanism arises from the reduction of grain size under stress in solid-state creep mechanisms which have grainsize-dependent viscosities (i.e., grain reduction leads to zones of weakness which concentrate deformation, increasing stress and thus further reducing grain size and enhancing weakening); this mechanism has been proposed to be the cause for mylonitic shear zones and is thought to be a basic ingredient of crustal and lithospheric deformation (Kameyama et al. 1997; Jin et al. 1998; Braun et al. 1999). Finally, material that undergoes brittle or combined brittle-ductile 1529-6466/00/0051-0013$05.00





     

deformation can, with large strain, experience concentration of microcrack and void populations developing weak bands on which deformation focuses, leading to further cracking, weakening, and invariably shear localization. This phenomenon occurs in the process of dilatant plasticity in metals as well as crustal rocks (Lemonds and Needleman 1986; Ashby and Sammis 1990; Lemaitre 1992; Hansen and Schreyer 1992; Lockner 1995; Mathur et al. 1996; Krajcinovic 1996; Lyakhovsky et al. 1997; Regenauer-Lieb 1999). The inference that much of the lithosphere undergoes combined brittle-ductile behavior (Kohlstedt et al. 1995; Evans and Kohlstedt 1995), combined with the longevity of plate boundaries (Gurnis et al. 2000), have motivated some (Bercovici 1998; Tackley 1998; Bercovici et al. 2001a,b; Bercovici and Ricard 2003; Auth et al. 2002) to consider that the primary weakening mechanism that leads to plate boundary formation is due to damage, that is, semi-ductile void and microcrack formation. In this paper we examine in some mathematical detail the fundamental physics of shearlocalizing feedback mechanisms by using the basic model of a simple-shear layer. Such a layer could represent anything from a strike-slip shear zone to a subducting slab to a glacial gravity flow, etc. Simple one-dimensional shear-flows have been considered many times before, primarily for viscous-heating and thermal-runaway effects (e.g., Yuen and Schubert 1977; Schubert and Yuen 1978; Fleitout and Froidevaux 1980). Here, we will use this model to examine three basic feedback mechanisms thought relevant to the mantle and lithosphere, i.e., (1) viscous heating with temperature-dependent viscosity; (2) a simple damage formulation to model semiductile-semibrittle behavior; and (3) grainsizedependent rheology and recrystallization. (While we focus on only three basic mechanisms, a more complete survey of many shear-localizing mechanisms, but with less mathematical detail, is provided by Regenauer-Lieb and Yuen (2002).) An important aspect of any feedback mechanism is not only how a rheology-controlling macroscopic or microscopic property is enhanced by deformation, but also how that property is restored, lost, equilibrated or “healed”; thus we will also consider some basic healing mechanisms. After our foray into simply theory, we will discuss recent progress on some of the more complicated aspects of these feedback mechanisms, and will close with a brief summary. THEORETICAL PRELIMINARIES To illustrate the essential physics of shear localization with various rheological mechanisms, we consider a model of unidirectional simple-shear flow in an infinitely long channel. The channel is infinite in the  and  directions and is  wide, with boundaries at  "!# that move in opposite directions along the  direction. Since the channel is uniform and infinite in  and  , none of the properties of the medium (stress, velocity, temperature, etc.) are functions of  and  , but only of  and time $ . In this case, the general force balance equation (i.e., assuming slow creeping flow and thus negligible acceleration terms) is

%

'&)(+*-,.(/10 , 23

(1)

where * is pressure or isotropic stress, 0 the deviatoric stress tensor, 2 the density and 3 the gravitational acceleration vector. The component of this equation in the  direction, given that there is no dependence on  and  (assuming that 3 is only in the  direction)

Analysis of Shear Localization in Lithosphere becomes

% 



  





(2)

where is the shear stress that is acting on surfaces facing in  the  direction but pulling  on them in the  direction. The above equation implies that which is a constant in  . For viscous flows, shear localization is measured in terms of strain-rate,

    ( , ( 

(3)

where is the strain-rate tensor and  the velocity vector. In our example     where  is the velocity in the  direction and  the unit vector in the  direction; thus the  #%$ . strain-rate tensor has only one unique component " !  

#

We make the simplifying assumption that stress is related to strain-rate plus only one other variable which we’ll denote for now by & which is either temperature, void fraction or grainsize. The constitutive relation between stress and strain rate is thus

0  ('  &*)+" !   , , ( ./ +/ (4) 



where ' is viscosity, " !   0  (which is related to the 2nd strain-rate invariant), and , is a material property which includes the bulk viscosity. In our example, since  1  
  there is ony one stress and one strain-rate and our constitutive law becomes



 ('  2 & )+" ! 3" !  where " ! 4 #%$ . #

(5)

If the relation between stress and strain-rate is monotonic (i.e., an increase in one always causes an increase in the other), and there is no feedback between strain-rate and the state variable & , then there can be no localization in our example. In particular, say we employ a power-law rheology, then '  6 5 "+! 798;: : where 5 is a constant and < is the power-law index. For pseudo-plastics (like many food products such as mustard and ketchup, and silicates in the dislocation-creep regime) =1? so that viscosity decreases  with increasing strain-rate. However, stress and thus resistance to motion, or strength,  5 " ! A@CB always increases " !   +D 5  B is a with strain-rate, even if the viscosity decreases. Moreover, in our example,  constant across the layer because and 5 are constants, and thus without any heterogeneity in strain-rate, there is no shear-localization anywhere in the layer. For the model system to develop shear-localization, it must allow non-constant strainrate across the layer (the localization is a region of high strain-rate surrounded by low-strain rate). This can be accomplished in our model if the effective stress–strain-rate relation is non-monotonic (i.e., there is a maximum or minimum stress), such that one stress can yield at least two-possible strain-rates. As we show below, such an effective rheology generally arises from a feedback between the state variable & and the stress or strain-rate. SHEAR LOCALIZING FEEDBACK MECHANISMS One requirement of a feedback mechanism that permits shear-localization is that the evolution of & must depend on the deformation, i.e., strain or strain-rate of the system; in





     

our examples we will focus on completely irrecoverable deformation and thus work with strain-rate (i.e., we will take a viscous-flow approach). Thus, for a feedback mechanism to exist, an increase in strain-rate must lead to a change in & that subsequently reduces the viscosity, causing (through Eqn. 5) yet larger strain-rate, and so on. We will examine our three basic feedback mechanisms (thermal, simple-damage and grainsize controlled) for simple decay-type loss mechanisms that facilitate the analysis. The decay-loss mechanism, however, is not always realistic in the case of thermally-controlled, and possibly even damage-controlled feedback mechanisms, wherein diffusive loss is as important; therefore, we will analyze these feedbacks for diffusive loss as well, which leads to different mathematical solutions but the interpretation is related to the simpler decay-loss cases. Thermal feedback with decay-loss healing In considering the feedback between viscous heating and temperature-dependent viscosity, we define & to be temperature  . The evolution equation for  involving temporal change, advective transport, decay loss and a deformation-related source term is

 

2





$

 ,- /( '&    &   , ('   ) " !  " !

(6)

where 2 is density,  is heat capacity,  is a heat-transfer coefficient, and  is the value of  to which the unforced system equilibrates (e.g., it could be the temperature of heat reservoirs in contact with the boundaries of the channel). A heat-transfer coefficient canD be used instead of diffusive loss if the layer thickness  is very small (in which case   where  is thermal conductivity). The last term on the right represents the deformational work, which for simple single-phase systems goes into dissipative heating and thus an increase in internal/thermal energy (i.e., via an increase in entropy). As discussed in the next subsection, in multi-phase systems, part of this work can also go into creation of surface energy by, say, creating new microcracks, voids and/or defects. Given the nature of our simple one-dimensional system, Equation (6) becomes



2 



$

  &  .&   , ('   )" ! 3" !

(7)

Let us first consider steady-state solutions such that Equation (7) becomes

    &   ('   )+" ! " !

(8)

To solve this system we must specify the viscosity law for '   )+" !  . Materials that undergo solid-state creep (diffusion or dislocation creep) have a temperature-dependence of viscosity given by an Arrhenius factor; for our 1-D system the viscosity law can be written in as

'   5:  " ! 97 8;: :

(9)

where 5 is constant,  is the activation enthalpy, the gas constant and < the powerlaw index; for Newtonian diffusion creep < 4? , while typical values for non-Newtonian dislocation creep in rocks is < "! to # . We can reduce the number of free parameters by defining variables in terms of the natural scales of the system. We define  $% ,& ')( and " !  " ! ' "! * where  ' and " ! ' are as-yet-unknown scales, and ( and "+! * are dimensionless

Analysis of Shear Localization in Lithosphere




temperature and strain-rate, respectively. Substituting these expressions into Equation (8) while using Equation (9) leads to (dropping primes for simplicity)





 ' 7 (        :   " ! 7 : : (10) "  7 6  : 5  ! '  : D where (    ' . Likewise, the constitutive law Equation (5)–along with Equation (9)– becomes



7         :   " ! : 7 " 5  ! '  A@CB

(11)

The expressions in square brackets are dimensionless groups, and we can choose  ' and " ! ' to make two of these go to unity. In particular, we define

 ' 





"!' 



: : 7   5

(12)

such that Equation (10) and Equation (11) become our steady-state governing equations:

7 7 (#    " ! B ! :

(13)

"

and

7    7  " ! : 

(14)

respectively, where

"







5 B  



:# 7 7

(15)

Although we have reduced the number of free parameters, our system still depends on the dimensionless boundary or background temperature ( and the power-law index < . Note D %*) %% that since we have chosen  %$%&(' ? K as our temperature scale (i.e., +$# kJ/mol and and ,$,- J/K/mol), then our typical backround lithospheric nondimensional temper%/.  . We will, however, also examine higher background temperature ( will be of order ? atures" for the sake of comparison. Equation (13) and Equation (14) lead to a parametric equation for stress and strain-rate " (i.e., and " ! are functions of each other through the “parameter” ( )

"

7 0 (   

7 : 7 )

" ! 0 ( B  .  7 

7 :# 7

(16)

which is thus an effective constitutive law. Curves for the effective constitutive relations for versus " ! for various < and ( are shown in Figure 1. In this case a variety of behavior is available. For large ( one can obtain a simple power-law like behavior wherein stress monotonically increases with strain-rate and thus no shear-localization is possible. For relatively low ( a non-monotonic effective stress–strain-rate constitutive relation occurs which therefore permits multiple strain-rates across the layer and the possibility of shear





     

3

3

thermal feedback (n varied)

thermal feedback ( θο varied)

n=1 2

θο = 0.01

2

n=3 n=5

1

θο = 0.2

1

θο = 1

stress

θο = 0.01 0 −0.1

0.5

n=3

.

0.2

0.5

0.8

1.1

1.4

1.7

2.0

0 −0.1

.

0.2

5

simple damage feedback

0.5

0.8

1.1

1.4

1.7

2.0

grain−size effect

0.4 4 0.3

m=3 3

0.2

m=2

2 0.1

m=1 0.0

.

0

n=p=3

1 5

10

15

0

10

20

30

strain−rate Figure 1: Effective “stress versus strain-rate” constitutive laws for three different feedback mechanisms with decay-type loss, as discussed in the text. The top two panels show the thermal feedback, i.e., where the sheared layer undergoes viscous heating that then influences the viscosity because of its temperature dependence. In these cases viscosity is possibly also a function of strain-rate through a power-law dependence that is controlled by the power-law index  (i.e., if  there is no dependence on strain-rate, and for   viscosity decreases with increasing strain-rate). The top left panel considers the thermal feedback with the background or boundary temperature  fixed at the value indicated, and for three different power-law indices  . The top right panel shows the same feedback for different background temperatures and one power-law index. For low enough background temperatures there is a region at low strain-rates where stress rapidly increases and then decreases with increasing strain-rate, which implies a resistance build-up and strength loss akin to stick-slip behavior. For yet higher strain-rates stress begins to increase again indicating a strength recovery; however, the rate of recovery with increasing strain-rate is much less for larger power-law indices  . For high background temperatures, there is no stick-slip-type region. The bottom two panels show feedback mechanisms for simple damage (deformation increases void/microcrack density which in turn reduces viscosity) and grainsize effects (stress reduces grain size which reduces viscosity). The damage mechanism gives a very simple stick-slip-type behavior with no strength-recovery. The grainsize effect is dependent on power-law index  , and indices that control the grainsize dependence of viscosity (if  viscosity is grainsize independent) and the grainsize dependence of grain growth  (if  grain-growth is grainsize independent). Stick-slip behavior in the grainsize feedback mechanism occurs with no finite resistance build-up; i.e., resistance is effectively infinite when strain-rate is infinitesimally small. For increasing strainrate there is strength loss and recovery also. However, the stick-slip type behavior only exists for sufficient grainsize dependence of viscosity, i.e.,  .

Analysis of Shear Localization in Lithosphere



localization. Indeed, in these low- ( cases, as many as three possible strain-rates can occur for a given stress (i.e., the constitutive curves have both a local maximum and a minimum). The constitutive curves that involve multiple strain-rates for a given stress are also considered to have a pseudo-stick-slip behavior. In particular, at low strain-rates the material resistance is large and climbs rapidly with increased deformation-rate, what would be analogous to “stick” behavior. At a peak stress, however, stress and resistance give way to the ”slip” behavior and loss of resistance or strength with increased strain-rate (the region of the constitutive curve with negative slope). In the thermal feedback, stress eventually begins to climb again at high strain-rates, albeit very gradually, as is characteristic of deformation in a hot, low-viscosity medium (which of course, at these high strain-rates and dissipative heating, it is). Such strength loss and stick-slip behavior is often considered closely associated with shear-localization. The multiplicity of states (which then leads to mulitple strain-rates) in this system is illustrated by considering the possible solutions to temperature ( in the steady-state system; these are of course the roots to Equation (13) and Equation (14), which can be combined into one transcendental equation "

7 (    

B 

"

(17)

These roots as functions of stress are shown in Figure 2. Three roots are possible, and % these represent a cold, highly viscous one (near (  ), an intermediate-temperature one and a hot, low-viscosity one.


  

There are, however, several constraints on the existence of these three roots. For a D   7  has two local extrema (a local ? , the function ( background temperature ( D maximum and minimum), and for ( = ? it has no local extrema (only a monotonic D increase with ( ). Thus, for ( ? there are three possible solutions to EquationD (17), but for larger ( there is only one solution. Nevertheless, the temperature scale  is so largeD that essentially all realistic lithospheric temperatures will satisfy the constraint of ( ? .

 



"

There is a further constraint on the possibility of shear localization in that the three " only exist for B beneath a maximum, as shown in Figure 2, which is itself" possible roots % . ?  , the maximum stress a function of ( . For plausible lithospheric values of (  % is given by B $ ? (see Fig. 2). The scale factor by which we redimensionalize is given by Equation (15); we can estimate this factor using typical silicate values for the % . given constants, namely <  ! , 5 B   # ' ? see Turcotte and Schubert D D   , (e.g., % ) %/. ) . . 1982),  " %$%& ' ? K (see above), and     $ ' ? ) assuming % %.  MPa, and the critical  1? km. With the resulting scale factor, which is of the order ?  % value of $ ! ' ? (having used <  ! ) we find the maximum is of the order of % MPa. Thus essentially all tectonic stresses will fall below this maximum stress and ? thus permit multiple solutions and thus shear localization.





 





 



  

However, because steady-state solutions to these equations exist does not mean they are stable against perturbations. Although we could perform a linear stability analysis of these solutions, or even calculate a full nonlinear time-dependent solution, it is instructive to D inspect the evolution equation. In particular, if we nondimensionalize time by 2   , then





      7

1.0

thermal feedback

simple damage

0.8 3

θ 10

φ

log

0.6

.

−1

θο = 0.01

0.4

θο = 0.03

θο = 0.5

−5

0.2

−9

.

0.0 −8

0

8

16

24

log τ

32

40

0.0

0.1

0.2

n+1

τ

0.3

0.4

0.5

10

2.0

grain−size effect

1.8

m=1 m=2

α

1.6

m=3 1.4

1.2

n=p=3 1.0 1.20

2.40

3.60

τ

4.80

6.00

Figure 2: Solutions of the state variables (temperature , void density  and grainsize  ) as functions of stress  for the same three feedback mechanisms of Figure 1. For each mechanism at least two possible solutions exist for a single stress  . Since multiple possible states exist for one stress across the sheared layer, then mulitple viscosities and strain-rates can exist within the layer also, which is a condition for shearlocalization (since it implies nonhomogenous shear). For the thermal feedback mechanism, three possible solutions exist for stresses beneath a maximum  which is itself a function of background temperature  and power-law index  (note  is absorbed into the abscissa of the plot). The damage mechanism allows two possible solutions for a given stress beneath a maximum stress    . For the grainsize mechanism, however, two possible solutions exist if the grainsize dependent of viscosity is sufficient (i.e.,  ) and such multiple solutions exist only for a stress above a minimum value, instead of beneath a maximum as with the other two feedback mechanisms.

along with the other " nondimensionalization steps discussed above, Equation (7) becomes





(

$

 &( ,

B 

.

 7 

(18)

If we plot # versus ( (Fig. 3) – a plot often referred to in dynamical systems as a phase % # diagram– we can see that the steady solutions occur where the curves intersect the #  # line. The stability of either steady solution is given by the the slope of the curve at the relevant interesection. If the slope is negative at the steady solution then it is stable; i.e., solutions on the curve to either side of the steady one will evolve to the steady solution (in particular, if the solution is to the left of the steady case with a smaller ( then its growth rate is positive and it will grow toward the steady value; if it is to the right with a larger ( , the growth rate is negative and the solution will decay to the steady value). In contrast if the slope is positive at the steady solution then that solution is unstable (solutions to either side of the steady one will diverge from it). In the case of Equation (18) (Fig. 3) we see




Analysis of Shear Localization in Lithosphere 0.6

thermal feedback

1.5

thermal feedback

1.0

τ = 1.4

d θ /dt

τ = 1.4 .

0

0.5

τ = 1.2

0.0

n=3 θο = 1

θο = 0.01

−0.6

1.0

1

.

n=3

τ=1 0

τ = 1.2 τ=1

θ

2

3

−0.5 0.00

1.75

1

damage feedback

3.50

θ

τ=2

grain−size effect

τ = 0.6 τ=3

d α/dt

d φ /dt

0

τ = 0.4 0

.

−1

τ = 0.2

τ=4

n=m=p=3

T −0.7 0.0

0.5

φ

1.0

−2 1.0

1.5

α

2.0

Figure 3: Phase diagrams for the same three feedback mechanisms of Figures 1 and 2, which show time-rate of change of the state variables versus the state variables themselves (i.e., temperature , void density  and grainsize  ). All panels show phase trajectories for different values of stress  as indicated; the top two panels show the thermal feedback with two different background temperatures  . These phase diagrams illustrate the temporal behavior of the feedback mechanisms and the stability of the steady solutions illustrated in Figure 2. Steady solutions occur where the phase trajectory crosses the line where the time-rate of change is zero. The trajectory slope at these intersections indicates the stability of the steady solution; i.e., the solution is unstable if the slope is positive, and stable if negative (see text for further discussion). Trajectories that do not cross the zero-rate line have no steady solution. See text for further discussion.

that if the background temperature ( is sufficiently small (such that there are three possible D steady solutions) then both the cold/viscous and the hot/low-viscosity cases are stable ( (  $ vs. ( has a negative slope at that root), while the intermediate-temperature case is unstable. Therefore unsteady solutions in the vicinity of the intermediate-temperature root will evolve toward one or the other stable root; temperatures hotter than the intermediate case will grow to the hot steady state, and colder temperatures will go to the cold case. Therefore, since the layer will in the end have two different temperatures and two strainrates, then weak zones and localization can be expected to occur (as long as the initial temperature is in the vicinity of the intermediate root). However, when a localized weak zone develops, it cannot weaken and localize without bound but must stop at the hot/lowviscosity solution. Thus a fault-like shear localization is impossible for this mechanism. D % ) $ & ' ? K both the However, it should be noted that given the temperature scale  intermediate and hot/low-viscosity roots are at temperatures near or well in excess of the melting temperature. Therefore, this mechanism would in fact give way to a melting in-





     

stability well before these roots were reached. With such melting and extreme viscosity and strength drop, an extremely sharp localization is possible, although that is beyond the simple physics we are exploring in this model. Simple-damage feedback with decay-loss healing The large region of the lithosphere that is thought to undergo combined brittle-ductile behavior (Kohlstedt et al. 1995) is suggestive of a damage-induced shear localization whereby microcracks forming in zones of ductile deformation focus into weak zones that thereby concentrate deformation and accelerate their own nucleation and growth, in the end creating distinct shear bands that appear like faults. Damage-induced shear localization is also appealing given the various roles it offers for water, which is thought to be a facilitator of plate tectonics, and the cause for plate tectonics on Earth but not the other terrestrial planets (Tozer 1985). For example, volatiles such as water can either facilitate damage through hydro-fracturing, pore-pressure-reduction of friction and simple lubrication of slip zones; and/or the damaged material (with finite pore space and open grain boundaries) can facilitate hydrolytic weakening of minerals and/or pressure-solution creep (Gu´eguen and Palciauskas 1994; Ranalli 1995). Although the role of water has generally been thought to be confined to only the upper few tens of kilometers of the lithosphere, there is growing evidence that it is present and influencing Earth’s rheology and evolution at hundreds of kilometers depth (Hirth and Kohlstedt 1996; Karato 2002). A simplified viscous damage theory termed the “void-volatile” mechanism (Bercovici, 1998; see also Regenauer-Lieb, 1999) involves modelling the evolution of microcrack or void density with a scalar transport law in which deformational work provides the source for void creation (similar to the heat-transport equation with frictional heating included). In this case, the weakening variable & is set equivalent to void volume fraction (i.e., porosity)  , and we write a simple transport law which in our 1-D example is

  2 



$

 & 

&



 ,  ' 



 )" ! 3" !

(19)



where  is the proportionality factor between and energy per unit mass,  is analogous   to  , and is the background value of to which the unforced system equilibrates. In this equation we have assumed that some fraction  of the deformational work goes into creation of surface energy by creating new microcracks, voids and/or defects; this is often referred to as stored-work, or, as discussed in the 1920s by G.I. Taylor, through his experiments on torsional deformation of metals, as latent-heat of cold work (Farren and Taylor 1925; Taylor and Quinney 1934). Finally, we assume a linear dependence of viscosity on  , which is relevant for the sort of two-phase mixtures assumed here (Bercovici 1998; Ricard et al. 2001). In particular, if the mixture is of a solid matrix with viscosity ' and a void-filling fluid with viscosity ' , then simple mixture theory will yield an effective   viscosity of ' , '   ? &  , which can be recast as

'  '  ? & where '  variability.





(20)

' is the reference viscosity and 



D

 ' &1' 6 ' controls viscosity

Again we first consider steady solutions to Equation (19), i.e., solutions of

 



&



  '  ? &

 "   !

(21)

Analysis of Shear Localization in Lithosphere




To reduce the number of free parameters in the system, we again use the natural scales of the system by first defining









,

'

(22)



where ' is an as yet unknown porosity scale and  is a rescaled porosity perturbation. In this case we rewrite viscosity as



'  '

? &

'  ? &

where ' 



  '   ?  & 

(23)



 . At this point we can define the porosity scale as  D ?  &  % 

' 

(24)

such that our viscosity law becomes

'  ' ? &  

(25)

We define strain-rate also in terms of a dimensionless quantity, namely " !  " ! '" ! * where " ! ' is a dimensional scale and " ! * is a dimensionless variable. In this case, the constitutive law  &  (Eqn. 5 using ) leads to 



 (' " ! ' 

 ? &  3" ! *

(26)

and Equation (21) becomes 

  ? &      '  " ! '  

  ? &  3" ! *

(27)

where the terms in square brackets are dimensionless groups. We choose our strain-rate scale to set one of these groups to unity:

   ? &   '

"!' 

 '

(28)

Thus, our two governing equations, Equation (26) and Equation (27) become, after dropping the" primes,

 ? &  3" !

(29)

    ? &   "!

(30)

 and

where "





  ('   ? & 

(31)





     

represents a dimensionless imposed stress. With these equations we can eliminate  to arrive at" an effective constitutive law

"!

 ? , "! 

(32)

which is shown in Figure 1. As noted in the previous section, because stress increases sharply below a certain strain-rate (in this case at " ! ? ) and then decreases for higher strain-rates, this effective constitutive law has been referred to explicitly as a stick-slip rheology (Whitehead and Gans 1974; Bercovici 1993). However, in contrast to the thermal feedback mechanism, the material continues to lose strength (stress decreases) as strainrate goes to infinity; this implies that effectively infinite strain rates can be achieved, and the ongoing strength loss is indicative of a runaway effect, which will become more evident momentarily. " Apart from the effective constitutive law, the actual solutions for  and " ! in terms of are instructive for inferring shear-localization in the layer. The steady-state solutions require " "


"?

    ? !



&

 



"!  

?

"




? !

? &



 

(33)

and thus for a given there are two possible porosities and strain-rates, a high-porosity/fast (i.e., rapidly deforming) one associated with the , root, and a low-porosity/slow case associated with the & root (Fig. 2). Shear localization requires inhomogeneous strain-rate, i.e., for a given stress across the layer localization requires at least two possible strain-rates to co-exist, which this feedback mechanism also provides. Again, shear localization would be manifest if the layer contained some regions with the slow solution " and some with the fast solution (the localizations would then be the fast zones). Equation (33) shows that there are no steady-state solutions if =  , in which case there are only time-dependent solutions. We can infer both the nature of time-dependent solutions and the stability of steady-state solutions by inspectingD the evolution equation Equation (19). In particular, if we nondimensionalize time by 2  , then along with the " other nondimensionalization steps discussed above, Equation (19) becomes

 

 $

&  , 



 ? & 

(34)

If we plot # versus  (Fig. 3) we again see that the steady solutions occur where the curves # % intersect the #  line and the stability of steady solutions is given by the the slope of the # curve at the relevant interesection. In this case, the low-porosity/slow steady solution is stable and the high-porosity/fast one is unstable. However, this does not mean that the system will everywhere evolve to the slow solution; indeed, values of  even slightly greater than that for the fast solution will grow toward even larger porosities. Although the maximum  is 1 (lest we have a negative viscosity), solutions with porosities larger than that for the high-porosity/fast solution will undergo a run-away to   ? and hence zero viscosity. Indeed, by inspection of Equation (34), an urealistic solution of  infinitesimally larger than 1 leads to an infinitely large negative growth rate which forces such unrealisic solutions to decay back to   ? . In essence,   ? is a stable singularity. In those regions with   ?






Analysis" of Shear Localization in Lithosphere

D

the strain-rate (which goes as  ? &   ) will be infinite. However, the size of any zeroviscosity zone is of finite width since it is entirely dependent on the initial porosity field; it will be as large as any region whose initial porosity exceeded that for the high-porosity/fast steady solution and thus grew to   ? . Thus although the strain-rate might be " heterogeneous, the shear zone does not necessarily self-localize, although the infinite strain-rates D require the velocity drop across a zero-viscosity zone to be infinite. Finally, if = ?  , there are no possible steady solutions (see Eqn. 33 and Fig. 1) and the entire system goes to   ? , i.e., the entire layer undergoes a runaway and has an infinite velocity drop across it.

"

To put some of these dimensionless results in a physical perspective, we can try to redimensionalize the critical stress   (below which there is shear-localization since two strain-rate solutions are permitted). The simple damage theory is idealized but we can relate some of the constants in the scaling equation Equation (31) to physical parameters. First, the partitioning fraction  is likely to be of order 0.1 (Ricard and Bercovici 2002), and  is % 
likely to be near unity. We can assume a very high lithospheric viscosity of ' $ ? Pa s   (Beaumont 1976; Watts et al. 1982), and that background porosity ? . The quantitiy  D is more difficult to quantify; it has units that represent the loss of energy density (  ) per second. The energy density is likely given by surface energy on cracks, which D in terms of volumetric energy density will be surface energy  –typically of order 1-100  for crack surfaces (Jaeger and Cook 1979; Cooper and Kohlstedt 1982; Atkinson 1987; Atkinson and Meredith 1987)– times void or crack curvature  (the inverse of void or crack size which . is typically 100-1000 ). The loss rate of this energy is unclear but we will assume that is controlled by loss of void-filling material, namely water, through Darcy-type flow; in D   

 ' , where  2 this case the flux rate (Darcy velocity) of fluid is of the order % . )   % %% D is permeability (of order ? for porous rocks), 2 is water density ( ? ), % . Pa s). The time-scale for energy loss will be  is gravity, and ' is water viscosity ( ? D typically   D where  is the length-scale of the shear-zone (1-10 km), and thus  will scale as %  . Given these values, the critical dimensional stress from Equation (31) is of the order several 100 to 1000 MPa, which is indeed typical of or larger than tectonic stresses. Thus we can speculate that tectonic stresses are sufficient to induce localization with this damage mechanism.









Grainsize feedback Grainsize reduction is an important cause for shear localization. The association of shear localization and grainsize reduction is evident in mylonite shear zones (White et al. 1980; Jin et al. 1998). However, the link between grainsize reduction and shear localization is not clearly understood. The way in which grainsize is controlled during deformation is more complicated than other variables such as temperature. In fact, some recent analyses underscored the difficulties in explaining shear localization through grainsize reduction (Bresser et al. 2001; Mont´esi and Zuber 2002), suggesting a need for more complete analysis of this problem. This section reviews some key concepts in grainsize sensitivity of deformation with the emphasis on the processes of dynamic recrystallization. Grainsize controls the rheology through two contrasting ways. Under low temperature conditions where dislocation recovery and diffusion do not occur, grainboundaries act as a barrier for dislocation motion and the presence of many grainboundaries causes hardening (the Hall-Petch effect; see Cottrell 1953). In contrast, when diffusion and resultant dislo-

     




cation recovery become effective (at high temperature  , i.e., exceeding roughly 50% of the melting temperature), then a smaller grainsize leads to weakening because deformation by diffusive mass transport occurs more efficiently for smaller grain sizes. Under most tectonic conditions except the shallow lithosphere, diffusive mass transfer is likely to be important, and we consider the effects of diffusion creep here. For simplicity, we assume (as in the previous sections) that viscosity depends only on grainsize and strain-rate, i.e.,

'  '  )+" !   '






  " 97 8;: : ! "

!




(35)

%

  & ! and <  ? for diffusion creep, and $ where and <  ! & # for dislo" cation creep;  is a reference grainsize and ! is the reference strain-rate, both of which are defined below. The rheological law Equation (35) prescribes dependence of viscosity on grainsize and strain-rate, although strictly speaking these effects do not occur simultaneously since grainsize dependence occurs in diffusion creep, and strain-rate dependence occurs in dislocation creep. Moreover, one of the dominant processes leading to grainsize reduction, namely dynamic recrystallization, occurs only in the dislocation-creep regime, and thus the feedback between grainsize reduction (while in the dislocation-creep regime) and grainsize-weakening (while in the diffusion-creep regime) only occurs if the material essentially experiences both creep regimes. This would seem to constrain the feedback mechanism to only be effective near the transition between creep regimes. However, the actual rheology of the material is complicated because it is essentially composed of two phases, i.e., coarse-grained strong matrix (which deforms by dislocation creep) and finegrained weak regions (deforming by diffusion creep). The flow law parameters such as < and in such a two-phase material are not well known but are likely to be between those for dislocation and diffusion creep. Therefore when deformation reduces the grainsize, there is a feedback mechanism that could causes self-weakening and shear localization.




There are several processes by which grainsize evolves. The first is dynamic recrystallization wherein gradients in dislocation density drive formation of new grainboundaries and/or mobilization of existing boundaries leading to the modification of grainsize. Evidence of dynamic recrystallization is ubiquitous in naturally deformed rocks (Urai et al. 1986). The second process involves chemical reactions, including phase transformations (Furusho and Kanagawa 1999; Karato et al. 2001; Rubie 1983, 1984; St¨unitz and Tullis 2001). In both cases, the kinetics of nucleation and growth controls the grainsize during deformation or chemical reactions, whereas after the completion of these processes, grain growth, through the reduction of grain-boundary surface energy, plays an important role when the grainsize is small enough. Whether grainsize modification causes shear localization depends on the competition between the weakening effect due to grainsize reduction and the hardening effect due to graingrowth, the details of which are specific to the particular processes involved. Here we consider the effects of grainsize evolution during dynamic recrystallization. In this case, and in the context of our simple-shear flow model, grainsize evolution may be described by

 

$



"!

& " 



&   ,





 .



(36)

(Karato 1996; Kameyama et al. 1997) where the first term represents the effect of dynamic recrystallization on grainsize and the second term the effect of graingrowth; " $ ? is

Analysis of Shear Localization in Lithosphere






the strain needed to complete dynamic recrystallization (Karato et al. 1980); K is a rate constant for graingrowth;  & ! such that grain growth is fastest for smaller grains when grain-boundary surface energy is large and diminishes as grains increase in size; and

is the equilibrium value of if there were no grain-growth and is in effect the grainsize at which dynamic recrystallization is completed. This grainsize may be determined by the nucleation-growth and impingement of growing subgrains on one another (Derby and Ashby 1987), or by subgrain rotation and dislodgement under deformation (Shimizu 1998). The implicit assumptions behind Equation (36) are, first, that grainsize reduction and graingrowth occurs simultaneously, and, second, that the time scale to obtain a shortterm equilibrium grainsize is significantly smaller than the time scale for graingrowth (therefore  is assumed to be a constant). As in previous sections, we nondimensionalize Equation (36) by defining -   $  $ * , and " !   " ! * and thus obtain (dropping the primes for simplicity)




,






$

 & "!





& ?6 ,

" the constitutive law Similarly,



(37)

  ' " ! becomes, using Equation (35),

 " ! A@CB




.

where "





(38)



" (' 



(39)

In steady state, Equation (37) leads to


"! 




. (40)

& ?

in which" case we can re-write Equation (38) as








 .  A@CB B 
& ?

(41) "

We can therefore display an effective constitutive law, as we did previously, by writing as a function of " ! parametrically " through
, as shown in Figure 1. There is no steadystate for
1? where grainsize increases both by dynamic recrystallization and by graingrowth % (assuming that both the imposed and the resulting " ! are = ). We will thus consider only the case
= ? , that is when an initially large grainsize is reduced by deformation.




For sufficiently large (i.e., sufficient grainsize dependence), the effective constitutive law is similar to that for the thermal-feedback case in which there is a region at low strain-rates whereby stress decreases with strain-rate (the “slip” regions), and a region at higher-strain rates whereby stress increases gradually. The stress-loss region is, as discussed before, an apparent feature of shear-localizing mechanisms. In contrast to the other feedback mechanisms, however, there is no finite region at low-strain-rates where stress increases sharply with strain-rate. Instead the peak strength occurs at zero strain-rate, which is similar to true stick-slip behavior whereby the “stick” response involves no motion until




     

a peak or yield stress is achieved, after which there is progressive loss of strength (i.e., failure ensues). The stick-slip region only exists when there is a minimum stress in the effective consti" (see Fig. " 1) which means that somewere on the curve tutive curve






  "! %






 "!